id sid tid token lemma pos ajst-31990 1 1 1 1 NUM ajst-31990 1 2 a a DET ajst-31990 1 3 comparison comparison NOUN ajst-31990 1 4 of of ADP ajst-31990 1 5 machine machine NOUN ajst-31990 1 6 learning learning NOUN ajst-31990 1 7 - - PUNCT ajst-31990 1 8 based base VERB ajst-31990 1 9 classification classification NOUN ajst-31990 1 10 and and CCONJ ajst-31990 1 11 recognition recognition NOUN ajst-31990 1 12 methods method NOUN ajst-31990 1 13 for for ADP ajst-31990 1 14 parkinson parkinson NOUN ajst-31990 1 15 's 's PART ajst-31990 1 16 prediction prediction NOUN ajst-31990 1 17 jiayu jiayu PROPN ajst-31990 1 18 zhang zhang PROPN ajst-31990 1 19 school school PROPN ajst-31990 1 20 of of ADP ajst-31990 1 21 international international ADJ ajst-31990 1 22 education education NOUN ajst-31990 1 23 , , PUNCT ajst-31990 1 24 guangdong guangdong PROPN ajst-31990 1 25 university university PROPN ajst-31990 1 26 of of ADP ajst-31990 1 27 technology technology PROPN ajst-31990 1 28 , , PUNCT ajst-31990 1 29 guangzhou guangzhou PROPN ajst-31990 1 30 , , PUNCT ajst-31990 1 31 china china PROPN ajst-31990 1 32 3223009955@mail2.gdut.edu.cn 3223009955@mail2.gdut.edu.cn PROPN ajst-31990 1 33 abstract abstract NOUN ajst-31990 1 34 . . PUNCT ajst-31990 2 1 parkinson parkinson NOUN ajst-31990 2 2 's 's PART ajst-31990 2 3 disease disease NOUN ajst-31990 2 4 ( ( PUNCT ajst-31990 2 5 pd pd NOUN ajst-31990 2 6 ) ) PUNCT ajst-31990 2 7 is be AUX ajst-31990 2 8 a a DET ajst-31990 2 9 neurodegenerative neurodegenerative ADJ ajst-31990 2 10 disease disease NOUN ajst-31990 2 11 that that PRON ajst-31990 2 12 is be AUX ajst-31990 2 13 often often ADV ajst-31990 2 14 associated associate VERB ajst-31990 2 15 with with ADP ajst-31990 2 16 abnormal abnormal ADJ ajst-31990 2 17 voice voice NOUN ajst-31990 2 18 function function NOUN ajst-31990 2 19 . . PUNCT ajst-31990 3 1 early early ADJ ajst-31990 3 2 recognition recognition NOUN ajst-31990 3 3 of of ADP ajst-31990 3 4 speech speech NOUN ajst-31990 3 5 features feature NOUN ajst-31990 3 6 is be AUX ajst-31990 3 7 crucial crucial ADJ ajst-31990 3 8 for for ADP ajst-31990 3 9 the the DET ajst-31990 3 10 diagnosis diagnosis NOUN ajst-31990 3 11 and and CCONJ ajst-31990 3 12 intervention intervention NOUN ajst-31990 3 13 of of ADP ajst-31990 3 14 pd pd PROPN ajst-31990 3 15 , , PUNCT ajst-31990 3 16 and and CCONJ ajst-31990 3 17 the the DET ajst-31990 3 18 development development NOUN ajst-31990 3 19 of of ADP ajst-31990 3 20 efficient efficient ADJ ajst-31990 3 21 and and CCONJ ajst-31990 3 22 reliable reliable ADJ ajst-31990 3 23 early early ADJ ajst-31990 3 24 screening screening NOUN ajst-31990 3 25 tools tool NOUN ajst-31990 3 26 has have AUX ajst-31990 3 27 become become VERB ajst-31990 3 28 a a DET ajst-31990 3 29 research research NOUN ajst-31990 3 30 hotspot hotspot NOUN ajst-31990 3 31 in in ADP ajst-31990 3 32 clinical clinical ADJ ajst-31990 3 33 medicine medicine NOUN ajst-31990 3 34 . . PUNCT ajst-31990 4 1 current current ADJ ajst-31990 4 2 research research NOUN ajst-31990 4 3 mostly mostly ADV ajst-31990 4 4 focuses focus VERB ajst-31990 4 5 on on ADP ajst-31990 4 6 single single ADJ ajst-31990 4 7 model model NOUN ajst-31990 4 8 optimization optimization NOUN ajst-31990 4 9 and and CCONJ ajst-31990 4 10 lacks lack VERB ajst-31990 4 11 a a DET ajst-31990 4 12 multidimensional multidimensional ADJ ajst-31990 4 13 assessment assessment NOUN ajst-31990 4 14 framework framework NOUN ajst-31990 4 15 . . PUNCT ajst-31990 5 1 this this DET ajst-31990 5 2 study study NOUN ajst-31990 5 3 compares compare VERB ajst-31990 5 4 the the DET ajst-31990 5 5 classification classification NOUN ajst-31990 5 6 performance performance NOUN ajst-31990 5 7 of of ADP ajst-31990 5 8 three three NUM ajst-31990 5 9 machine machine NOUN ajst-31990 5 10 learning learning NOUN ajst-31990 5 11 models model NOUN ajst-31990 5 12 , , PUNCT ajst-31990 5 13 logistic logistic ADJ ajst-31990 5 14 regression regression NOUN ajst-31990 5 15 , , PUNCT ajst-31990 5 16 xgboost xgboost ADV ajst-31990 5 17 and and CCONJ ajst-31990 5 18 support support VERB ajst-31990 5 19 vector vector NOUN ajst-31990 5 20 machine machine NOUN ajst-31990 5 21 ( ( PUNCT ajst-31990 5 22 svm svm PROPN ajst-31990 5 23 ) ) PUNCT ajst-31990 5 24 , , PUNCT ajst-31990 5 25 in in ADP ajst-31990 5 26 pd pd PROPN ajst-31990 5 27 prediction prediction NOUN ajst-31990 5 28 based base VERB ajst-31990 5 29 on on ADP ajst-31990 5 30 the the DET ajst-31990 5 31 publicly publicly ADV ajst-31990 5 32 available available ADJ ajst-31990 5 33 parkinson parkinson NOUN ajst-31990 5 34 's 's PART ajst-31990 5 35 speech speech NOUN ajst-31990 5 36 feature feature NOUN ajst-31990 5 37 dataset dataset VERB ajst-31990 5 38 from from ADP ajst-31990 5 39 uci uci PROPN ajst-31990 5 40 . . PUNCT ajst-31990 6 1 key key PROPN ajst-31990 6 2 features feature NOUN ajst-31990 6 3 are be AUX ajst-31990 6 4 extracted extract VERB ajst-31990 6 5 through through ADP ajst-31990 6 6 data datum NOUN ajst-31990 6 7 feature feature NOUN ajst-31990 6 8 engineering engineering NOUN ajst-31990 6 9 , , PUNCT ajst-31990 6 10 oversampling oversample VERB ajst-31990 6 11 is be AUX ajst-31990 6 12 used use VERB ajst-31990 6 13 to to PART ajst-31990 6 14 address address VERB ajst-31990 6 15 data datum NOUN ajst-31990 6 16 imbalance imbalance NOUN ajst-31990 6 17 , , PUNCT ajst-31990 6 18 and and CCONJ ajst-31990 6 19 metrics metric NOUN ajst-31990 6 20 such such ADJ ajst-31990 6 21 as as ADP ajst-31990 6 22 confusion confusion NOUN ajst-31990 6 23 matrices matrix NOUN ajst-31990 6 24 are be AUX ajst-31990 6 25 used use VERB ajst-31990 6 26 for for ADP ajst-31990 6 27 multidimensional multidimensional ADJ ajst-31990 6 28 assessment assessment NOUN ajst-31990 6 29 . . PUNCT ajst-31990 7 1 the the DET ajst-31990 7 2 results result NOUN ajst-31990 7 3 of of ADP ajst-31990 7 4 the the DET ajst-31990 7 5 study study NOUN ajst-31990 7 6 showed show VERB ajst-31990 7 7 that that SCONJ ajst-31990 7 8 logistic logistic ADJ ajst-31990 7 9 regression regression NOUN ajst-31990 7 10 had have VERB ajst-31990 7 11 the the DET ajst-31990 7 12 best good ADJ ajst-31990 7 13 overall overall ADJ ajst-31990 7 14 performance performance NOUN ajst-31990 7 15 on on ADP ajst-31990 7 16 the the DET ajst-31990 7 17 validation validation NOUN ajst-31990 7 18 set set NOUN ajst-31990 7 19 , , PUNCT ajst-31990 7 20 with with ADP ajst-31990 7 21 an an DET ajst-31990 7 22 accuracy accuracy NOUN ajst-31990 7 23 of of ADP ajst-31990 7 24 81.47 81.47 NUM ajst-31990 7 25 % % NOUN ajst-31990 7 26 and and CCONJ ajst-31990 7 27 a a DET ajst-31990 7 28 recall recall NOUN ajst-31990 7 29 rate rate NOUN ajst-31990 7 30 of of ADP ajst-31990 7 31 95 95 NUM ajst-31990 7 32 % % NOUN ajst-31990 7 33 for for ADP ajst-31990 7 34 healthy healthy ADJ ajst-31990 7 35 people people NOUN ajst-31990 7 36 , , PUNCT ajst-31990 7 37 but but CCONJ ajst-31990 7 38 its its PRON ajst-31990 7 39 patient patient NOUN ajst-31990 7 40 recall recall NOUN ajst-31990 7 41 rate rate NOUN ajst-31990 7 42 ( ( PUNCT ajst-31990 7 43 71 71 NUM ajst-31990 7 44 % % NOUN ajst-31990 7 45 ) ) PUNCT ajst-31990 7 46 still still ADV ajst-31990 7 47 needs need VERB ajst-31990 7 48 to to PART ajst-31990 7 49 be be AUX ajst-31990 7 50 optimized optimize VERB ajst-31990 7 51 . . PUNCT ajst-31990 8 1 xgboost xgboost PROPN ajst-31990 8 2 does do AUX ajst-31990 8 3 not not PART ajst-31990 8 4 adequately adequately ADV ajst-31990 8 5 deal deal VERB ajst-31990 8 6 with with ADP ajst-31990 8 7 class class NOUN ajst-31990 8 8 imbalance imbalance NOUN ajst-31990 8 9 , , PUNCT ajst-31990 8 10 under under NOUN ajst-31990 8 11 - - PUNCT ajst-31990 8 12 recognizes recognize VERB ajst-31990 8 13 minority minority NOUN ajst-31990 8 14 classes class NOUN ajst-31990 8 15 , , PUNCT ajst-31990 8 16 and and CCONJ ajst-31990 8 17 svm svm ADJ ajst-31990 8 18 performs perform VERB ajst-31990 8 19 poorly poorly ADV ajst-31990 8 20 with with ADP ajst-31990 8 21 the the DET ajst-31990 8 22 current current ADJ ajst-31990 8 23 data datum NOUN ajst-31990 8 24 distribution distribution NOUN ajst-31990 8 25 . . PUNCT ajst-31990 9 1 this this DET ajst-31990 9 2 study study NOUN ajst-31990 9 3 provides provide VERB ajst-31990 9 4 a a DET ajst-31990 9 5 theoretical theoretical ADJ ajst-31990 9 6 basis basis NOUN ajst-31990 9 7 for for ADP ajst-31990 9 8 model model NOUN ajst-31990 9 9 selection selection NOUN ajst-31990 9 10 for for ADP ajst-31990 9 11 disease disease NOUN ajst-31990 9 12 diagnosis diagnosis NOUN ajst-31990 9 13 and and CCONJ ajst-31990 9 14 lays lay VERB ajst-31990 9 15 down down ADP ajst-31990 9 16 a a DET ajst-31990 9 17 methodology methodology NOUN ajst-31990 9 18 for for ADP ajst-31990 9 19 intelligent intelligent ADJ ajst-31990 9 20 processing processing NOUN ajst-31990 9 21 of of ADP ajst-31990 9 22 multimodal multimodal ADJ ajst-31990 9 23 medical medical ADJ ajst-31990 9 24 data datum NOUN ajst-31990 9 25 . . PUNCT ajst-31990 10 1 keywords keyword NOUN ajst-31990 10 2 : : PUNCT ajst-31990 10 3 phonetic phonetic ADJ ajst-31990 10 4 character character NOUN ajst-31990 10 5 analysis analysis NOUN ajst-31990 10 6 ; ; PUNCT ajst-31990 10 7 parkinson parkinson NOUN ajst-31990 10 8 's 's PART ajst-31990 10 9 disease disease NOUN ajst-31990 10 10 prediction prediction NOUN ajst-31990 10 11 ; ; PUNCT ajst-31990 10 12 logistic logistic ADJ ajst-31990 10 13 regression regression NOUN ajst-31990 10 14 ; ; PUNCT ajst-31990 10 15 xgboost xgboost ADV ajst-31990 10 16 ; ; PUNCT ajst-31990 10 17 support support NOUN ajst-31990 10 18 vector vector NOUN ajst-31990 10 19 machines machine NOUN ajst-31990 10 20 . . PUNCT ajst-31990 11 1 1 1 X ajst-31990 11 2 . . X ajst-31990 11 3 introduction introduction NOUN ajst-31990 11 4 parkinson parkinson NOUN ajst-31990 11 5 's 's PART ajst-31990 11 6 disease disease NOUN ajst-31990 11 7 ( ( PUNCT ajst-31990 11 8 pd pd NOUN ajst-31990 11 9 ) ) PUNCT ajst-31990 11 10 is be AUX ajst-31990 11 11 a a DET ajst-31990 11 12 degenerative degenerative ADJ ajst-31990 11 13 disease disease NOUN ajst-31990 11 14 of of ADP ajst-31990 11 15 the the DET ajst-31990 11 16 nervous nervous ADJ ajst-31990 11 17 system system NOUN ajst-31990 11 18 , , PUNCT ajst-31990 11 19 with with ADP ajst-31990 11 20 typical typical ADJ ajst-31990 11 21 symptoms symptom NOUN ajst-31990 11 22 including include VERB ajst-31990 11 23 muscle muscle NOUN ajst-31990 11 24 tremors tremor NOUN ajst-31990 11 25 , , PUNCT ajst-31990 11 26 slowed slow VERB ajst-31990 11 27 movements movement NOUN ajst-31990 11 28 and and CCONJ ajst-31990 11 29 unsteady unsteady ADJ ajst-31990 11 30 gait gait NOUN ajst-31990 11 31 . . PUNCT ajst-31990 12 1 as as SCONJ ajst-31990 12 2 the the DET ajst-31990 12 3 disease disease NOUN ajst-31990 12 4 progresses progress VERB ajst-31990 12 5 , , PUNCT ajst-31990 12 6 patients patient NOUN ajst-31990 12 7 are be AUX ajst-31990 12 8 often often ADV ajst-31990 12 9 accompanied accompany VERB ajst-31990 12 10 by by ADP ajst-31990 12 11 abnormalities abnormality NOUN ajst-31990 12 12 in in ADP ajst-31990 12 13 voice voice NOUN ajst-31990 12 14 function function NOUN ajst-31990 12 15 , , PUNCT ajst-31990 12 16 such such ADJ ajst-31990 12 17 as as ADP ajst-31990 12 18 reduced reduced ADJ ajst-31990 12 19 volume volume NOUN ajst-31990 12 20 , , PUNCT ajst-31990 12 21 monotone monotone ADJ ajst-31990 12 22 and and CCONJ ajst-31990 12 23 slurred slurred ADJ ajst-31990 12 24 speech speech NOUN ajst-31990 12 25 . . PUNCT ajst-31990 13 1 studies study NOUN ajst-31990 13 2 have have AUX ajst-31990 13 3 shown show VERB ajst-31990 13 4 that that SCONJ ajst-31990 13 5 early early ADJ ajst-31990 13 6 identification identification NOUN ajst-31990 13 7 of of ADP ajst-31990 13 8 changes change NOUN ajst-31990 13 9 in in ADP ajst-31990 13 10 speech speech NOUN ajst-31990 13 11 characteristics characteristic NOUN ajst-31990 13 12 is be AUX ajst-31990 13 13 crucial crucial ADJ ajst-31990 13 14 for for ADP ajst-31990 13 15 the the DET ajst-31990 13 16 diagnosis diagnosis NOUN ajst-31990 13 17 and and CCONJ ajst-31990 13 18 intervention intervention NOUN ajst-31990 13 19 of of ADP ajst-31990 13 20 pd pd PROPN ajst-31990 13 21 , , PUNCT ajst-31990 13 22 and and CCONJ ajst-31990 13 23 the the DET ajst-31990 13 24 development development NOUN ajst-31990 13 25 of of ADP ajst-31990 13 26 efficient efficient ADJ ajst-31990 13 27 and and CCONJ ajst-31990 13 28 reliable reliable ADJ ajst-31990 13 29 early early ADJ ajst-31990 13 30 screening screening NOUN ajst-31990 13 31 tools tool NOUN ajst-31990 13 32 has have AUX ajst-31990 13 33 become become VERB ajst-31990 13 34 a a DET ajst-31990 13 35 research research NOUN ajst-31990 13 36 hotspot hotspot NOUN ajst-31990 13 37 in in ADP ajst-31990 13 38 clinical clinical ADJ ajst-31990 13 39 medicine medicine NOUN ajst-31990 13 40 . . PUNCT ajst-31990 14 1 in in ADP ajst-31990 14 2 recent recent ADJ ajst-31990 14 3 years year NOUN ajst-31990 14 4 , , PUNCT ajst-31990 14 5 the the DET ajst-31990 14 6 combination combination NOUN ajst-31990 14 7 of of ADP ajst-31990 14 8 speech speech NOUN ajst-31990 14 9 feature feature NOUN ajst-31990 14 10 processing processing NOUN ajst-31990 14 11 and and CCONJ ajst-31990 14 12 machine machine NOUN ajst-31990 14 13 learning learning NOUN ajst-31990 14 14 techniques technique NOUN ajst-31990 14 15 has have AUX ajst-31990 14 16 provided provide VERB ajst-31990 14 17 new new ADJ ajst-31990 14 18 ideas idea NOUN ajst-31990 14 19 for for ADP ajst-31990 14 20 disease disease NOUN ajst-31990 14 21 diagnosis diagnosis NOUN ajst-31990 14 22 . . PUNCT ajst-31990 15 1 pd pd PROPN ajst-31990 15 2 as as ADP ajst-31990 15 3 a a DET ajst-31990 15 4 global global ADJ ajst-31990 15 5 health health NOUN ajst-31990 15 6 challenge challenge NOUN ajst-31990 15 7 was be AUX ajst-31990 15 8 also also ADV ajst-31990 15 9 mentioned mention VERB ajst-31990 15 10 in in ADP ajst-31990 15 11 the the DET ajst-31990 15 12 study study NOUN ajst-31990 15 13 by by ADP ajst-31990 15 14 neto neto PROPN ajst-31990 15 15 et et PROPN ajst-31990 16 1 al al PROPN ajst-31990 16 2 . . PROPN ajst-31990 17 1 traditional traditional ADJ ajst-31990 17 2 diagnostic diagnostic ADJ ajst-31990 17 3 methods method NOUN ajst-31990 17 4 rely rely VERB ajst-31990 17 5 on on ADP ajst-31990 17 6 physical physical ADJ ajst-31990 17 7 symptom symptom NOUN ajst-31990 17 8 assessment assessment NOUN ajst-31990 17 9 , , PUNCT ajst-31990 17 10 which which PRON ajst-31990 17 11 often often ADV ajst-31990 17 12 leads lead VERB ajst-31990 17 13 to to ADP ajst-31990 17 14 early early ADJ ajst-31990 17 15 missed miss VERB ajst-31990 17 16 diagnoses diagnosis NOUN ajst-31990 17 17 . . PUNCT ajst-31990 18 1 sound sound ADJ ajst-31990 18 2 analysis analysis NOUN ajst-31990 18 3 , , PUNCT ajst-31990 18 4 as as ADP ajst-31990 18 5 a a DET ajst-31990 18 6 non non ADJ ajst-31990 18 7 - - ADJ ajst-31990 18 8 invasive invasive ADJ ajst-31990 18 9 and and CCONJ ajst-31990 18 10 easily easily ADV ajst-31990 18 11 accessible accessible ADJ ajst-31990 18 12 tool tool NOUN ajst-31990 18 13 , , PUNCT ajst-31990 18 14 can can AUX ajst-31990 18 15 be be AUX ajst-31990 18 16 combined combine VERB ajst-31990 18 17 with with ADP ajst-31990 18 18 machine machine NOUN ajst-31990 18 19 learning learning NOUN ajst-31990 18 20 ( ( PUNCT ajst-31990 18 21 ml ml NOUN ajst-31990 18 22 ) ) PUNCT ajst-31990 18 23 techniques technique NOUN ajst-31990 18 24 to to PART ajst-31990 18 25 comprehensively comprehensively ADV ajst-31990 18 26 assess assess VERB ajst-31990 18 27 the the DET ajst-31990 18 28 effectiveness effectiveness NOUN ajst-31990 18 29 of of ADP ajst-31990 18 30 sound sound ADJ ajst-31990 18 31 analysis analysis NOUN ajst-31990 18 32 in in ADP ajst-31990 18 33 the the DET ajst-31990 18 34 early early ADJ ajst-31990 18 35 diagnosis diagnosis NOUN ajst-31990 18 36 of of ADP ajst-31990 18 37 pd pd NOUN ajst-31990 19 1 [ [ X ajst-31990 19 2 1 1 NUM ajst-31990 19 3 ] ] PUNCT ajst-31990 19 4 . . PUNCT ajst-31990 20 1 speech speech NOUN ajst-31990 20 2 characteristics characteristic NOUN ajst-31990 20 3 in in ADP ajst-31990 20 4 pd pd PROPN ajst-31990 20 5 patients patient NOUN ajst-31990 20 6 show show VERB ajst-31990 20 7 significant significant ADJ ajst-31990 20 8 differences difference NOUN ajst-31990 20 9 due due ADP ajst-31990 20 10 to to ADP ajst-31990 20 11 abnormalities abnormality NOUN ajst-31990 20 12 in in ADP ajst-31990 20 13 neuromuscular neuromuscular ADJ ajst-31990 20 14 control control NOUN ajst-31990 20 15 , , PUNCT ajst-31990 20 16 and and CCONJ ajst-31990 20 17 these these DET ajst-31990 20 18 acoustic acoustic ADJ ajst-31990 20 19 features feature NOUN ajst-31990 20 20 in in ADP ajst-31990 20 21 combination combination NOUN ajst-31990 20 22 with with ADP ajst-31990 20 23 nonlinear nonlinear ADJ ajst-31990 20 24 kinetic kinetic ADJ ajst-31990 20 25 parameters parameter NOUN ajst-31990 20 26 can can AUX ajst-31990 20 27 be be AUX ajst-31990 20 28 effective effective ADJ ajst-31990 20 29 in in ADP ajst-31990 20 30 characterising characterise VERB ajst-31990 20 31 the the DET ajst-31990 20 32 pathological pathological ADJ ajst-31990 20 33 state state NOUN ajst-31990 20 34 . . PUNCT ajst-31990 21 1 solana solana PROPN ajst-31990 21 2 - - PUNCT ajst-31990 21 3 lavalle lavalle PROPN ajst-31990 21 4 et et PROPN ajst-31990 21 5 al al PROPN ajst-31990 21 6 . . PROPN ajst-31990 21 7 developed develop VERB ajst-31990 21 8 a a DET ajst-31990 21 9 tool tool NOUN ajst-31990 21 10 for for ADP ajst-31990 21 11 assisted assist VERB ajst-31990 21 12 detection detection NOUN ajst-31990 21 13 of of ADP ajst-31990 21 14 pd pd PROPN ajst-31990 21 15 based base VERB ajst-31990 21 16 on on ADP ajst-31990 21 17 speech speech NOUN ajst-31990 21 18 analysis analysis NOUN ajst-31990 21 19 , , PUNCT ajst-31990 21 20 focusing focus VERB ajst-31990 21 21 on on ADP ajst-31990 21 22 the the DET ajst-31990 21 23 clinical clinical ADJ ajst-31990 21 24 interpretability interpretability NOUN ajst-31990 21 25 of of ADP ajst-31990 21 26 the the DET ajst-31990 21 27 algorithm algorithm NOUN ajst-31990 21 28 output output NOUN ajst-31990 21 29 , , PUNCT ajst-31990 21 30 i.e. i.e. X ajst-31990 21 31 , , PUNCT ajst-31990 21 32 providing provide VERB ajst-31990 21 33 not not PART ajst-31990 21 34 only only ADV ajst-31990 21 35 binary binary ADJ ajst-31990 21 36 classification classification NOUN ajst-31990 21 37 results result NOUN ajst-31990 21 38 , , PUNCT ajst-31990 21 39 but but CCONJ ajst-31990 21 40 also also ADV ajst-31990 21 41 contextual contextual ADJ ajst-31990 21 42 correlation correlation NOUN ajst-31990 21 43 analyses analysis NOUN ajst-31990 21 44 of of ADP ajst-31990 21 45 the the DET ajst-31990 21 46 relevant relevant ADJ ajst-31990 21 47 speech speech NOUN ajst-31990 21 48 features feature NOUN ajst-31990 21 49 , , PUNCT ajst-31990 21 50 to to PART ajst-31990 21 51 help help VERB ajst-31990 21 52 physicians physician NOUN ajst-31990 21 53 understand understand VERB ajst-31990 21 54 the the DET ajst-31990 21 55 basis basis NOUN ajst-31990 21 56 of of ADP ajst-31990 21 57 the the DET ajst-31990 21 58 diagnosis diagnosis NOUN ajst-31990 21 59 [ [ X ajst-31990 21 60 2 2 NUM ajst-31990 21 61 ] ] PUNCT ajst-31990 21 62 . . PUNCT ajst-31990 22 1 current current ADJ ajst-31990 22 2 research research NOUN ajst-31990 22 3 mostly mostly ADV ajst-31990 22 4 focuses focus VERB ajst-31990 22 5 on on ADP ajst-31990 22 6 the the DET ajst-31990 22 7 optimisation optimisation NOUN ajst-31990 22 8 of of ADP ajst-31990 22 9 a a DET ajst-31990 22 10 single single ADJ ajst-31990 22 11 model model NOUN ajst-31990 22 12 or or CCONJ ajst-31990 22 13 feature feature NOUN ajst-31990 22 14 and and CCONJ ajst-31990 22 15 lacks lack VERB ajst-31990 22 16 a a DET ajst-31990 22 17 multidimensional multidimensional ADJ ajst-31990 22 18 assessment assessment NOUN ajst-31990 22 19 framework framework NOUN ajst-31990 22 20 . . PUNCT ajst-31990 23 1 a a DET ajst-31990 23 2 study study NOUN ajst-31990 23 3 by by ADP ajst-31990 23 4 pah pah PROPN ajst-31990 23 5 et et PROPN ajst-31990 23 6 al al PROPN ajst-31990 23 7 . . PROPN ajst-31990 23 8 confirmed confirm VERB ajst-31990 23 9 the the DET ajst-31990 23 10 potential potential NOUN ajst-31990 23 11 of of ADP ajst-31990 23 12 acoustic acoustic ADJ ajst-31990 23 13 features feature NOUN ajst-31990 23 14 for for ADP ajst-31990 23 15 non non ADJ ajst-31990 23 16 - - ADJ ajst-31990 23 17 invasive invasive ADJ ajst-31990 23 18 diagnosis diagnosis NOUN ajst-31990 23 19 of of ADP ajst-31990 23 20 pd pd PROPN ajst-31990 23 21 ( ( PUNCT ajst-31990 23 22 svm svm VERB ajst-31990 23 23 binary binary ADJ ajst-31990 23 24 classification classification NOUN ajst-31990 23 25 accuracy accuracy NOUN ajst-31990 23 26 of of ADP ajst-31990 23 27 80.60 80.60 NUM ajst-31990 23 28 % % NOUN ajst-31990 23 29 ) ) PUNCT ajst-31990 23 30 , , PUNCT ajst-31990 23 31 but but CCONJ ajst-31990 23 32 the the DET ajst-31990 23 33 multivariate multivariate NOUN ajst-31990 23 34 classification classification NOUN ajst-31990 23 35 f1 f1 NOUN ajst-31990 23 36 - - PUNCT ajst-31990 23 37 score score NOUN ajst-31990 23 38 was be AUX ajst-31990 23 39 only only ADV ajst-31990 23 40 40.46 40.46 NUM ajst-31990 23 41 % % NOUN ajst-31990 23 42 revealing revealing ADJ ajst-31990 23 43 limitations limitation NOUN ajst-31990 23 44 in in ADP ajst-31990 23 45 multi multi ADJ ajst-31990 23 46 - - ADJ ajst-31990 23 47 disease disease ADJ ajst-31990 23 48 discrimination discrimination NOUN ajst-31990 23 49 , , PUNCT ajst-31990 23 50 and and CCONJ ajst-31990 23 51 optimised optimise VERB ajst-31990 23 52 feature feature NOUN ajst-31990 23 53 fusion fusion NOUN ajst-31990 23 54 and and CCONJ ajst-31990 23 55 classifier classifier NOUN ajst-31990 23 56 design design NOUN ajst-31990 23 57 is be AUX ajst-31990 23 58 needed need VERB ajst-31990 23 59 to to PART ajst-31990 23 60 improve improve VERB ajst-31990 23 61 performance performance NOUN ajst-31990 23 62 2 2 NUM ajst-31990 23 63 in in ADP ajst-31990 23 64 complex complex ADJ ajst-31990 23 65 scenarios scenario NOUN ajst-31990 23 66 [ [ X ajst-31990 23 67 3 3 NUM ajst-31990 23 68 ] ] PUNCT ajst-31990 23 69 . . PUNCT ajst-31990 24 1 the the DET ajst-31990 24 2 study study NOUN ajst-31990 24 3 by by ADP ajst-31990 24 4 sakar sakar PROPN ajst-31990 24 5 et et PROPN ajst-31990 24 6 al al PROPN ajst-31990 24 7 . . PROPN ajst-31990 24 8 used use VERB ajst-31990 24 9 tunable tunable ADJ ajst-31990 24 10 q q ADJ ajst-31990 24 11 - - PUNCT ajst-31990 24 12 factor factor NOUN ajst-31990 24 13 wavelet wavelet NOUN ajst-31990 24 14 transform transform NOUN ajst-31990 24 15 ( ( PUNCT ajst-31990 24 16 tqwt tqwt NOUN ajst-31990 24 17 ) ) PUNCT ajst-31990 24 18 as as ADP ajst-31990 24 19 the the DET ajst-31990 24 20 main main ADJ ajst-31990 24 21 feature feature NOUN ajst-31990 24 22 extraction extraction NOUN ajst-31990 24 23 method method NOUN ajst-31990 24 24 suitable suitable ADJ ajst-31990 24 25 for for ADP ajst-31990 24 26 high high ADJ ajst-31990 24 27 q q ADJ ajst-31990 24 28 - - PUNCT ajst-31990 24 29 factor factor NOUN ajst-31990 24 30 transform transform NOUN ajst-31990 24 31 of of ADP ajst-31990 24 32 speech speech NOUN ajst-31990 24 33 signals signal NOUN ajst-31990 24 34 , , PUNCT ajst-31990 24 35 applied apply VERB ajst-31990 24 36 to to ADP ajst-31990 24 37 pd pd NOUN ajst-31990 24 38 classification classification NOUN ajst-31990 24 39 . . PUNCT ajst-31990 25 1 overall overall ADV ajst-31990 25 2 , , PUNCT ajst-31990 25 3 the the DET ajst-31990 25 4 tqwt tqwt NOUN ajst-31990 25 5 method method NOUN ajst-31990 25 6 significantly significantly ADV ajst-31990 25 7 improves improve VERB ajst-31990 25 8 the the DET ajst-31990 25 9 classification classification NOUN ajst-31990 25 10 performance performance NOUN ajst-31990 25 11 [ [ X ajst-31990 25 12 4 4 NUM ajst-31990 25 13 ] ] PUNCT ajst-31990 25 14 . . PUNCT ajst-31990 26 1 lahmiri lahmiri PROPN ajst-31990 26 2 et et PROPN ajst-31990 26 3 al al PROPN ajst-31990 26 4 . . PROPN ajst-31990 26 5 showed show VERB ajst-31990 26 6 that that SCONJ ajst-31990 26 7 svm svm PROPN ajst-31990 26 8 combined combine VERB ajst-31990 26 9 with with ADP ajst-31990 26 10 feature feature NOUN ajst-31990 26 11 ranking rank VERB ajst-31990 26 12 techniques technique NOUN ajst-31990 26 13 were be AUX ajst-31990 26 14 able able ADJ ajst-31990 26 15 to to PART ajst-31990 26 16 effectively effectively ADV ajst-31990 26 17 differentiate differentiate VERB ajst-31990 26 18 between between ADP ajst-31990 26 19 pd pd NOUN ajst-31990 26 20 patients patient NOUN ajst-31990 26 21 and and CCONJ ajst-31990 26 22 healthy healthy ADJ ajst-31990 26 23 controls control NOUN ajst-31990 26 24 , , PUNCT ajst-31990 26 25 demonstrating demonstrate VERB ajst-31990 26 26 high high ADJ ajst-31990 26 27 sensitivity sensitivity NOUN ajst-31990 26 28 and and CCONJ ajst-31990 26 29 specificity specificity NOUN ajst-31990 26 30 ( ( PUNCT ajst-31990 26 31 99.63 99.63 NUM ajst-31990 26 32 % % NOUN ajst-31990 26 33 sensitivity sensitivity NOUN ajst-31990 26 34 and and CCONJ ajst-31990 26 35 82.79 82.79 NUM ajst-31990 26 36 % % NOUN ajst-31990 26 37 specificity specificity NOUN ajst-31990 26 38 ) ) PUNCT ajst-31990 26 39 , , PUNCT ajst-31990 26 40 and and CCONJ ajst-31990 26 41 that that SCONJ ajst-31990 26 42 the the DET ajst-31990 26 43 accuracy accuracy NOUN ajst-31990 26 44 of of ADP ajst-31990 26 45 the the DET ajst-31990 26 46 system system NOUN ajst-31990 26 47 based base VERB ajst-31990 26 48 on on ADP ajst-31990 26 49 speech speech NOUN ajst-31990 26 50 features feature NOUN ajst-31990 26 51 was be AUX ajst-31990 26 52 better well ADJ ajst-31990 26 53 than than ADP ajst-31990 26 54 that that PRON ajst-31990 26 55 of of ADP ajst-31990 26 56 the the DET ajst-31990 26 57 pd pd PROPN ajst-31990 26 58 detection detection NOUN ajst-31990 26 59 system system NOUN ajst-31990 26 60 based base VERB ajst-31990 26 61 on on ADP ajst-31990 26 62 mri mri NOUN ajst-31990 26 63 , , PUNCT ajst-31990 26 64 emotion emotion NOUN ajst-31990 26 65 and and CCONJ ajst-31990 26 66 handwriting handwriting NOUN ajst-31990 26 67 features feature NOUN ajst-31990 26 68 [ [ X ajst-31990 26 69 5 5 NUM ajst-31990 26 70 ] ] PUNCT ajst-31990 26 71 . . PUNCT ajst-31990 27 1 however however ADV ajst-31990 27 2 , , PUNCT ajst-31990 27 3 the the DET ajst-31990 27 4 characteristics characteristic NOUN ajst-31990 27 5 of of ADP ajst-31990 27 6 high high ADJ ajst-31990 27 7 - - PUNCT ajst-31990 27 8 dimensional dimensional ADJ ajst-31990 27 9 , , PUNCT ajst-31990 27 10 small small ADJ ajst-31990 27 11 - - PUNCT ajst-31990 27 12 sample sample NOUN ajst-31990 27 13 data datum NOUN ajst-31990 27 14 pose pose VERB ajst-31990 27 15 challenges challenge NOUN ajst-31990 27 16 to to ADP ajst-31990 27 17 classification classification NOUN ajst-31990 27 18 models model NOUN ajst-31990 27 19 . . PUNCT ajst-31990 28 1 tai tai PROPN ajst-31990 28 2 et et PROPN ajst-31990 28 3 al al PROPN ajst-31990 28 4 . . PROPN ajst-31990 28 5 proposed propose VERB ajst-31990 28 6 a a DET ajst-31990 28 7 speech speech NOUN ajst-31990 28 8 analytics analytic NOUN ajst-31990 28 9 - - PUNCT ajst-31990 28 10 based base VERB ajst-31990 28 11 approach approach NOUN ajst-31990 28 12 to to PART ajst-31990 28 13 recognise recognise VERB ajst-31990 28 14 patterns pattern NOUN ajst-31990 28 15 of of ADP ajst-31990 28 16 pd pd NOUN ajst-31990 28 17 by by ADP ajst-31990 28 18 pre pre VERB ajst-31990 28 19 - - ADJ ajst-31990 28 20 processing process VERB ajst-31990 28 21 the the DET ajst-31990 28 22 dataset dataset NOUN ajst-31990 28 23 for for ADP ajst-31990 28 24 feature feature NOUN ajst-31990 28 25 extraction extraction NOUN ajst-31990 28 26 and and CCONJ ajst-31990 28 27 dimensionality dimensionality NOUN ajst-31990 28 28 reduction reduction NOUN ajst-31990 28 29 of of ADP ajst-31990 28 30 the the DET ajst-31990 28 31 data datum NOUN ajst-31990 28 32 and and CCONJ ajst-31990 28 33 employing employ VERB ajst-31990 28 34 various various ADJ ajst-31990 28 35 machine machine NOUN ajst-31990 28 36 learning learning NOUN ajst-31990 28 37 models model NOUN ajst-31990 28 38 for for ADP ajst-31990 28 39 classification classification NOUN ajst-31990 28 40 , , PUNCT ajst-31990 28 41 where where SCONJ ajst-31990 28 42 four four NUM ajst-31990 28 43 supervised supervised ADJ ajst-31990 28 44 learning learning NOUN ajst-31990 28 45 algorithms algorithm NOUN ajst-31990 28 46 were be AUX ajst-31990 28 47 used use VERB ajst-31990 28 48 for for ADP ajst-31990 28 49 classification classification NOUN ajst-31990 28 50 : : PUNCT ajst-31990 28 51 multi multi ADJ ajst-31990 28 52 - - ADJ ajst-31990 28 53 layer layer ADJ ajst-31990 28 54 perceptron perceptron NOUN ajst-31990 28 55 ( ( PUNCT ajst-31990 28 56 mlp mlp PROPN ajst-31990 28 57 ) ) PUNCT ajst-31990 28 58 , , PUNCT ajst-31990 28 59 random random ADJ ajst-31990 28 60 forest forest NOUN ajst-31990 28 61 ( ( PUNCT ajst-31990 28 62 rf rf NOUN ajst-31990 28 63 ) ) PUNCT ajst-31990 28 64 , , PUNCT ajst-31990 28 65 logistic logistic ADJ ajst-31990 28 66 regression regression NOUN ajst-31990 28 67 ( ( PUNCT ajst-31990 28 68 lr lr NOUN ajst-31990 28 69 ) ) PUNCT ajst-31990 28 70 and and CCONJ ajst-31990 28 71 support support VERB ajst-31990 28 72 vector vector NOUN ajst-31990 28 73 machines machine NOUN ajst-31990 28 74 ( ( PUNCT ajst-31990 28 75 svm svm PROPN ajst-31990 28 76 ) ) PUNCT ajst-31990 28 77 so so SCONJ ajst-31990 28 78 as as SCONJ ajst-31990 28 79 to to PART ajst-31990 28 80 evaluate evaluate VERB ajst-31990 28 81 the the DET ajst-31990 28 82 model model NOUN ajst-31990 28 83 performance performance NOUN ajst-31990 28 84 comparatively comparatively ADV ajst-31990 28 85 [ [ X ajst-31990 28 86 6 6 NUM ajst-31990 28 87 ] ] PUNCT ajst-31990 28 88 . . PUNCT ajst-31990 29 1 the the DET ajst-31990 29 2 study study NOUN ajst-31990 29 3 by by ADP ajst-31990 29 4 wroge wroge PROPN ajst-31990 29 5 et et PROPN ajst-31990 29 6 al al PROPN ajst-31990 29 7 . . PROPN ajst-31990 29 8 used use VERB ajst-31990 29 9 an an DET ajst-31990 29 10 iphone iphone NOUN ajst-31990 29 11 application application NOUN ajst-31990 29 12 to to PART ajst-31990 29 13 collect collect VERB ajst-31990 29 14 digital digital ADJ ajst-31990 29 15 biomarkers biomarker NOUN ajst-31990 29 16 and and CCONJ ajst-31990 29 17 health health NOUN ajst-31990 29 18 data datum NOUN ajst-31990 29 19 from from ADP ajst-31990 29 20 pd pd PROPN ajst-31990 29 21 patients patient NOUN ajst-31990 29 22 and and CCONJ ajst-31990 29 23 non non NOUN ajst-31990 29 24 - - NOUN ajst-31990 29 25 patients patient NOUN ajst-31990 29 26 to to PART ajst-31990 29 27 extract extract VERB ajst-31990 29 28 and and CCONJ ajst-31990 29 29 characterise characterise NOUN ajst-31990 29 30 speech speech NOUN ajst-31990 29 31 , , PUNCT ajst-31990 29 32 using use VERB ajst-31990 29 33 a a DET ajst-31990 29 34 variety variety NOUN ajst-31990 29 35 of of ADP ajst-31990 29 36 supervised supervised ADJ ajst-31990 29 37 classification classification NOUN ajst-31990 29 38 algorithms algorithm NOUN ajst-31990 29 39 including include VERB ajst-31990 29 40 svm svm PROPN ajst-31990 29 41 , , PUNCT ajst-31990 29 42 gradient gradient ADJ ajst-31990 29 43 boosted boost VERB ajst-31990 29 44 decision decision NOUN ajst-31990 29 45 trees tree NOUN ajst-31990 29 46 ( ( PUNCT ajst-31990 29 47 gbdt gbdt PROPN ajst-31990 29 48 ) ) PUNCT ajst-31990 29 49 , , PUNCT ajst-31990 29 50 as as ADV ajst-31990 29 51 well well ADV ajst-31990 29 52 as as ADP ajst-31990 29 53 optimising optimise VERB ajst-31990 29 54 the the DET ajst-31990 29 55 model model NOUN ajst-31990 29 56 based base VERB ajst-31990 29 57 on on ADP ajst-31990 29 58 the the DET ajst-31990 29 59 f1 f1 NOUN ajst-31990 29 60 score score NOUN ajst-31990 29 61 and and CCONJ ajst-31990 29 62 recall.the recall.the DET ajst-31990 29 63 results result NOUN ajst-31990 29 64 show show VERB ajst-31990 29 65 that that SCONJ ajst-31990 29 66 the the DET ajst-31990 29 67 machine machine NOUN ajst-31990 29 68 learning learning NOUN ajst-31990 29 69 model model NOUN ajst-31990 29 70 exhibits exhibit VERB ajst-31990 29 71 high high ADJ ajst-31990 29 72 accuracy accuracy NOUN ajst-31990 29 73 in in ADP ajst-31990 29 74 pd pd NOUN ajst-31990 29 75 diagnosis diagnosis NOUN ajst-31990 30 1 [ [ X ajst-31990 30 2 7 7 NUM ajst-31990 30 3 ] ] PUNCT ajst-31990 30 4 . . PUNCT ajst-31990 31 1 this this DET ajst-31990 31 2 study study NOUN ajst-31990 31 3 is be AUX ajst-31990 31 4 based base VERB ajst-31990 31 5 on on ADP ajst-31990 31 6 the the DET ajst-31990 31 7 uci uci PROPN ajst-31990 31 8 public public PROPN ajst-31990 31 9 dataset dataset NOUN ajst-31990 31 10 , , PUNCT ajst-31990 31 11 covering cover VERB ajst-31990 31 12 754 754 NUM ajst-31990 31 13 - - PUNCT ajst-31990 31 14 dimensional dimensional ADJ ajst-31990 31 15 acoustic acoustic ADJ ajst-31990 31 16 features feature NOUN ajst-31990 31 17 , , PUNCT ajst-31990 31 18 and and CCONJ ajst-31990 31 19 solves solve VERB ajst-31990 31 20 the the DET ajst-31990 31 21 problems problem NOUN ajst-31990 31 22 of of ADP ajst-31990 31 23 feature feature NOUN ajst-31990 31 24 redundancy redundancy NOUN ajst-31990 31 25 and and CCONJ ajst-31990 31 26 category category NOUN ajst-31990 31 27 imbalance imbalance NOUN ajst-31990 31 28 through through ADP ajst-31990 31 29 chi chi ADJ ajst-31990 31 30 - - PUNCT ajst-31990 31 31 square square NOUN ajst-31990 31 32 test test NOUN ajst-31990 31 33 and and CCONJ ajst-31990 31 34 oversampling oversample VERB ajst-31990 31 35 techniques technique NOUN ajst-31990 31 36 , , PUNCT ajst-31990 31 37 comparing compare VERB ajst-31990 31 38 the the DET ajst-31990 31 39 performances performance NOUN ajst-31990 31 40 of of ADP ajst-31990 31 41 logistic logistic ADJ ajst-31990 31 42 regression regression NOUN ajst-31990 31 43 , , PUNCT ajst-31990 31 44 xgboost xgboost ADV ajst-31990 31 45 , , PUNCT ajst-31990 31 46 and and CCONJ ajst-31990 31 47 svm svm VERB ajst-31990 31 48 in in ADP ajst-31990 31 49 parkinson parkinson NOUN ajst-31990 31 50 's 's PART ajst-31990 31 51 speech speech NOUN ajst-31990 31 52 classification classification NOUN ajst-31990 31 53 from from ADP ajst-31990 31 54 the the DET ajst-31990 31 55 perspectives perspective NOUN ajst-31990 31 56 of of ADP ajst-31990 31 57 computational computational ADJ ajst-31990 31 58 efficiency efficiency NOUN ajst-31990 31 59 , , PUNCT ajst-31990 31 60 risk risk NOUN ajst-31990 31 61 of of ADP ajst-31990 31 62 overfitting overfitte VERB ajst-31990 31 63 , , PUNCT ajst-31990 31 64 and and CCONJ ajst-31990 31 65 feature feature NOUN ajst-31990 31 66 sensitivity sensitivity NOUN ajst-31990 31 67 . . PUNCT ajst-31990 32 1 the the DET ajst-31990 32 2 results result NOUN ajst-31990 32 3 will will AUX ajst-31990 32 4 clarify clarify VERB ajst-31990 32 5 the the DET ajst-31990 32 6 scenarios scenario NOUN ajst-31990 32 7 in in ADP ajst-31990 32 8 which which PRON ajst-31990 32 9 different different ADJ ajst-31990 32 10 algorithms algorithm NOUN ajst-31990 32 11 are be AUX ajst-31990 32 12 advantageous advantageous ADJ ajst-31990 32 13 for for ADP ajst-31990 32 14 early early ADJ ajst-31990 32 15 screening screening NOUN ajst-31990 32 16 : : PUNCT ajst-31990 32 17 logistic logistic ADJ ajst-31990 32 18 regression regression NOUN ajst-31990 32 19 is be AUX ajst-31990 32 20 suitable suitable ADJ ajst-31990 32 21 for for ADP ajst-31990 32 22 rapid rapid ADJ ajst-31990 32 23 primary primary ADJ ajst-31990 32 24 screening screening NOUN ajst-31990 32 25 by by ADP ajst-31990 32 26 virtue virtue NOUN ajst-31990 32 27 of of ADP ajst-31990 32 28 its its PRON ajst-31990 32 29 interpretability interpretability NOUN ajst-31990 32 30 , , PUNCT ajst-31990 32 31 xgboost xgboost ADP ajst-31990 32 32 excels excel NOUN ajst-31990 32 33 in in ADP ajst-31990 32 34 complex complex ADJ ajst-31990 32 35 pattern pattern NOUN ajst-31990 32 36 recognition recognition NOUN ajst-31990 32 37 by by ADP ajst-31990 32 38 gradient gradient NOUN ajst-31990 32 39 boosting boost VERB ajst-31990 32 40 trees tree NOUN ajst-31990 32 41 , , PUNCT ajst-31990 32 42 and and CCONJ ajst-31990 32 43 svm svm ADJ ajst-31990 32 44 utilises utilise NOUN ajst-31990 32 45 the the DET ajst-31990 32 46 kernel kernel PROPN ajst-31990 32 47 function function VERB ajst-31990 32 48 to to PART ajst-31990 32 49 achieve achieve VERB ajst-31990 32 50 efficient efficient ADJ ajst-31990 32 51 segmentation segmentation NOUN ajst-31990 32 52 in in ADP ajst-31990 32 53 high high ADJ ajst-31990 32 54 - - PUNCT ajst-31990 32 55 dimensional dimensional ADJ ajst-31990 32 56 spaces space NOUN ajst-31990 32 57 , , PUNCT ajst-31990 32 58 which which PRON ajst-31990 32 59 needs need VERB ajst-31990 32 60 to to PART ajst-31990 32 61 be be AUX ajst-31990 32 62 combined combine VERB ajst-31990 32 63 with with ADP ajst-31990 32 64 dimensionality dimensionality NOUN ajst-31990 32 65 reduction reduction NOUN ajst-31990 32 66 techniques technique NOUN ajst-31990 32 67 to to PART ajst-31990 32 68 improve improve VERB ajst-31990 32 69 small small ADJ ajst-31990 32 70 - - PUNCT ajst-31990 32 71 sample sample NOUN ajst-31990 32 72 adaptability adaptability NOUN ajst-31990 32 73 . . PUNCT ajst-31990 33 1 systematic systematic ADJ ajst-31990 33 2 comparison comparison NOUN ajst-31990 33 3 of of ADP ajst-31990 33 4 the the DET ajst-31990 33 5 performance performance NOUN ajst-31990 33 6 differences difference NOUN ajst-31990 33 7 between between ADP ajst-31990 33 8 the the DET ajst-31990 33 9 three three NUM ajst-31990 33 10 types type NOUN ajst-31990 33 11 of of ADP ajst-31990 33 12 models model NOUN ajst-31990 33 13 can can AUX ajst-31990 33 14 provide provide VERB ajst-31990 33 15 a a DET ajst-31990 33 16 theoretical theoretical ADJ ajst-31990 33 17 basis basis NOUN ajst-31990 33 18 for for ADP ajst-31990 33 19 the the DET ajst-31990 33 20 optimisation optimisation NOUN ajst-31990 33 21 of of ADP ajst-31990 33 22 clinical clinical ADJ ajst-31990 33 23 diagnostic diagnostic ADJ ajst-31990 33 24 tools tool NOUN ajst-31990 33 25 . . PUNCT ajst-31990 34 1 this this DET ajst-31990 34 2 study study NOUN ajst-31990 34 3 theoretically theoretically ADV ajst-31990 34 4 reveals reveal VERB ajst-31990 34 5 the the DET ajst-31990 34 6 relationship relationship NOUN ajst-31990 34 7 between between ADP ajst-31990 34 8 high high ADV ajst-31990 34 9 - - PUNCT ajst-31990 34 10 dimensional dimensional ADJ ajst-31990 34 11 sparse sparse ADJ ajst-31990 34 12 features feature NOUN ajst-31990 34 13 and and CCONJ ajst-31990 34 14 model model NOUN ajst-31990 34 15 generalisation generalisation NOUN ajst-31990 34 16 ; ; PUNCT ajst-31990 34 17 technology technology NOUN ajst-31990 34 18 to to PART ajst-31990 34 19 build build VERB ajst-31990 34 20 a a DET ajst-31990 34 21 multi multi ADJ ajst-31990 34 22 - - ADJ ajst-31990 34 23 dimensional dimensional ADJ ajst-31990 34 24 assessment assessment NOUN ajst-31990 34 25 framework framework NOUN ajst-31990 34 26 to to PART ajst-31990 34 27 guide guide VERB ajst-31990 34 28 healthcare healthcare PROPN ajst-31990 34 29 ai ai VERB ajst-31990 34 30 deployment deployment NOUN ajst-31990 34 31 ; ; PUNCT ajst-31990 34 32 the the DET ajst-31990 34 33 application application NOUN ajst-31990 34 34 recommends recommend VERB ajst-31990 34 35 optimal optimal ADJ ajst-31990 34 36 diagnostic diagnostic ADJ ajst-31990 34 37 tools tool NOUN ajst-31990 34 38 based base VERB ajst-31990 34 39 on on ADP ajst-31990 34 40 the the DET ajst-31990 34 41 hardware hardware NOUN ajst-31990 34 42 conditions condition NOUN ajst-31990 34 43 at at ADP ajst-31990 34 44 the the DET ajst-31990 34 45 grassroots grassroot NOUN ajst-31990 34 46 level.the level.the DET ajst-31990 34 47 research research NOUN ajst-31990 34 48 provides provide VERB ajst-31990 34 49 both both DET ajst-31990 34 50 support support NOUN ajst-31990 34 51 for for ADP ajst-31990 34 52 clinical clinical ADJ ajst-31990 34 53 model model NOUN ajst-31990 34 54 selection selection NOUN ajst-31990 34 55 and and CCONJ ajst-31990 34 56 a a DET ajst-31990 34 57 methodological methodological ADJ ajst-31990 34 58 basis basis NOUN ajst-31990 34 59 for for ADP ajst-31990 34 60 multimodal multimodal ADJ ajst-31990 34 61 medical medical ADJ ajst-31990 34 62 data datum NOUN ajst-31990 34 63 processing processing NOUN ajst-31990 34 64 . . PUNCT ajst-31990 35 1 2 2 X ajst-31990 35 2 . . X ajst-31990 35 3 research research NOUN ajst-31990 35 4 methodology methodology NOUN ajst-31990 35 5 2.1 2.1 NUM ajst-31990 35 6 . . PUNCT ajst-31990 36 1 experimental experimental ADJ ajst-31990 36 2 environment environment NOUN ajst-31990 36 3 configuration configuration NOUN ajst-31990 36 4 the the DET ajst-31990 36 5 tool tool NOUN ajst-31990 36 6 platform platform NOUN ajst-31990 36 7 used use VERB ajst-31990 36 8 in in ADP ajst-31990 36 9 this this DET ajst-31990 36 10 research research NOUN ajst-31990 36 11 is be AUX ajst-31990 36 12 vscode vscode NOUN ajst-31990 36 13 with with ADP ajst-31990 36 14 python python NOUN ajst-31990 36 15 3.8 3.8 NUM ajst-31990 36 16 + + CCONJ ajst-31990 36 17 kernel kernel NOUN ajst-31990 36 18 , , PUNCT ajst-31990 36 19 where where SCONJ ajst-31990 36 20 anaconda anaconda PROPN ajst-31990 36 21 is be AUX ajst-31990 36 22 used use VERB ajst-31990 36 23 to to PART ajst-31990 36 24 build build VERB ajst-31990 36 25 the the DET ajst-31990 36 26 virtual virtual ADJ ajst-31990 36 27 environment environment NOUN ajst-31990 36 28 . . PUNCT ajst-31990 37 1 process process NOUN ajst-31990 37 2 and and CCONJ ajst-31990 37 3 visualise visualise VERB ajst-31990 37 4 data datum NOUN ajst-31990 37 5 using use VERB ajst-31990 37 6 libraries library NOUN ajst-31990 37 7 such such ADJ ajst-31990 37 8 as as ADP ajst-31990 37 9 numpy numpy NOUN ajst-31990 37 10 , , PUNCT ajst-31990 37 11 pandas panda NOUN ajst-31990 37 12 , , PUNCT ajst-31990 37 13 matplotlib matplotlib PROPN ajst-31990 37 14 and and CCONJ ajst-31990 37 15 seaborn seaborn PROPN ajst-31990 37 16 , , PUNCT ajst-31990 37 17 and and CCONJ ajst-31990 37 18 reference reference NOUN ajst-31990 37 19 machine machine NOUN ajst-31990 37 20 learning learning NOUN ajst-31990 37 21 models model NOUN ajst-31990 37 22 and and CCONJ ajst-31990 37 23 tools tool NOUN ajst-31990 37 24 using use VERB ajst-31990 37 25 libraries library NOUN ajst-31990 37 26 such such ADJ ajst-31990 37 27 as as ADP ajst-31990 37 28 scikit scikit NOUN ajst-31990 37 29 - - PUNCT ajst-31990 37 30 learn learn NOUN ajst-31990 37 31 . . PUNCT ajst-31990 38 1 2.2 2.2 NUM ajst-31990 38 2 . . PUNCT ajst-31990 38 3 research research NOUN ajst-31990 38 4 process process NOUN ajst-31990 38 5 drawing draw VERB ajst-31990 38 6 on on ADP ajst-31990 38 7 the the DET ajst-31990 38 8 research research NOUN ajst-31990 38 9 idea idea NOUN ajst-31990 38 10 proposed propose VERB ajst-31990 38 11 by by ADP ajst-31990 38 12 tai tai PROPN ajst-31990 38 13 et et PROPN ajst-31990 38 14 al al PROPN ajst-31990 38 15 . . PROPN ajst-31990 38 16 and and CCONJ ajst-31990 38 17 the the DET ajst-31990 38 18 research research NOUN ajst-31990 38 19 process process NOUN ajst-31990 38 20 of of ADP ajst-31990 38 21 wroge wroge PROPN ajst-31990 38 22 et et PROPN ajst-31990 38 23 al al PROPN ajst-31990 38 24 . . PUNCT ajst-31990 39 1 [ [ X ajst-31990 39 2 6,7 6,7 NUM ajst-31990 39 3 ] ] PUNCT ajst-31990 39 4 , , PUNCT ajst-31990 39 5 this this DET ajst-31990 39 6 study study NOUN ajst-31990 39 7 designed design VERB ajst-31990 39 8 an an DET ajst-31990 39 9 overall overall ADJ ajst-31990 39 10 flow flow NOUN ajst-31990 39 11 chart chart NOUN ajst-31990 39 12 for for ADP ajst-31990 39 13 this this DET ajst-31990 39 14 study study NOUN ajst-31990 39 15 , , PUNCT ajst-31990 39 16 as as SCONJ ajst-31990 39 17 shown show VERB ajst-31990 39 18 in in ADP ajst-31990 39 19 figure figure NOUN ajst-31990 39 20 1 1 NUM ajst-31990 39 21 . . SYM ajst-31990 39 22 3 3 NUM ajst-31990 39 23 figure figure NOUN ajst-31990 39 24 1 1 NUM ajst-31990 39 25 . . PUNCT ajst-31990 39 26 overall overall ADJ ajst-31990 39 27 flow flow NOUN ajst-31990 39 28 chart chart NOUN ajst-31990 39 29 of of ADP ajst-31990 39 30 this this DET ajst-31990 39 31 study study NOUN ajst-31990 39 32 ( ( PUNCT ajst-31990 39 33 picture picture NOUN ajst-31990 39 34 credit credit NOUN ajst-31990 39 35 : : PUNCT ajst-31990 39 36 original original ADJ ajst-31990 39 37 ) ) PUNCT ajst-31990 39 38 2.3 2.3 NUM ajst-31990 39 39 . . PUNCT ajst-31990 40 1 modelling model VERB ajst-31990 40 2 principles principle NOUN ajst-31990 40 3 2.3.1 2.3.1 NUM ajst-31990 40 4 logistic logistic ADJ ajst-31990 40 5 regression regression NOUN ajst-31990 40 6 . . PUNCT ajst-31990 41 1 the the DET ajst-31990 41 2 logistic logistic ADJ ajst-31990 41 3 regression regression NOUN ajst-31990 41 4 model model NOUN ajst-31990 41 5 associates associates PROPN ajst-31990 41 6 input input NOUN ajst-31990 41 7 features feature NOUN ajst-31990 41 8 with with ADP ajst-31990 41 9 category category NOUN ajst-31990 41 10 probabilities probability NOUN ajst-31990 41 11 through through ADP ajst-31990 41 12 a a DET ajst-31990 41 13 combination combination NOUN ajst-31990 41 14 of of ADP ajst-31990 41 15 linear linear ADJ ajst-31990 41 16 combinations combination NOUN ajst-31990 41 17 and and CCONJ ajst-31990 41 18 probabilistic probabilistic ADJ ajst-31990 41 19 mappings mapping NOUN ajst-31990 41 20 , , PUNCT ajst-31990 41 21 which which PRON ajst-31990 41 22 is be AUX ajst-31990 41 23 suitable suitable ADJ ajst-31990 41 24 for for ADP ajst-31990 41 25 binary binary ADJ ajst-31990 41 26 classification classification NOUN ajst-31990 41 27 problems problem NOUN ajst-31990 41 28 ( ( PUNCT ajst-31990 41 29 e.g. e.g. ADV ajst-31990 41 30 , , PUNCT ajst-31990 41 31 disease disease NOUN ajst-31990 41 32 prediction prediction NOUN ajst-31990 41 33 , , PUNCT ajst-31990 41 34 spam spam NOUN ajst-31990 41 35 detection detection NOUN ajst-31990 41 36 ) ) PUNCT ajst-31990 41 37 , , PUNCT ajst-31990 41 38 is be AUX ajst-31990 41 39 highly highly ADV ajst-31990 41 40 interpretable interpretable ADJ ajst-31990 41 41 , , PUNCT ajst-31990 41 42 and and CCONJ ajst-31990 41 43 is be AUX ajst-31990 41 44 suitable suitable ADJ ajst-31990 41 45 for for ADP ajst-31990 41 46 preliminary preliminary ADJ ajst-31990 41 47 feature feature NOUN ajst-31990 41 48 screening.it screening.it NUM ajst-31990 41 49 works work VERB ajst-31990 41 50 best well ADV ajst-31990 41 51 when when SCONJ ajst-31990 41 52 there there PRON ajst-31990 41 53 is be VERB ajst-31990 41 54 an an DET ajst-31990 41 55 approximate approximate ADJ ajst-31990 41 56 linear linear ADJ ajst-31990 41 57 relationship relationship NOUN ajst-31990 41 58 between between ADP ajst-31990 41 59 features feature NOUN ajst-31990 41 60 and and CCONJ ajst-31990 41 61 categories category NOUN ajst-31990 41 62 . . PUNCT ajst-31990 42 1 logistic logistic ADJ ajst-31990 42 2 regression regression NOUN ajst-31990 42 3 , , PUNCT ajst-31990 42 4 which which PRON ajst-31990 42 5 can can AUX ajst-31990 42 6 be be AUX ajst-31990 42 7 used use VERB ajst-31990 42 8 to to PART ajst-31990 42 9 estimate estimate VERB ajst-31990 42 10 the the DET ajst-31990 42 11 relationship relationship NOUN ajst-31990 42 12 between between ADP ajst-31990 42 13 one one NUM ajst-31990 42 14 or or CCONJ ajst-31990 42 15 more more ADV ajst-31990 42 16 independent independent ADJ ajst-31990 42 17 variables variable NOUN ajst-31990 42 18 and and CCONJ ajst-31990 42 19 a a DET ajst-31990 42 20 binary binary ADJ ajst-31990 42 21 ( ( PUNCT ajst-31990 42 22 dichotomous dichotomous ADJ ajst-31990 42 23 ) ) PUNCT ajst-31990 42 24 outcome outcome NOUN ajst-31990 42 25 variable variable NOUN ajst-31990 42 26 , , PUNCT ajst-31990 42 27 was be AUX ajst-31990 42 28 also also ADV ajst-31990 42 29 mentioned mention VERB ajst-31990 42 30 in in ADP ajst-31990 42 31 the the DET ajst-31990 42 32 study study NOUN ajst-31990 42 33 by by ADP ajst-31990 42 34 schober schober PROPN ajst-31990 42 35 et et PROPN ajst-31990 42 36 al al PROPN ajst-31990 42 37 . . PROPN ajst-31990 42 38 and and CCONJ ajst-31990 42 39 has have AUX ajst-31990 42 40 been be AUX ajst-31990 42 41 welcomed welcome VERB ajst-31990 42 42 in in ADP ajst-31990 42 43 healthcare healthcare PROPN ajst-31990 42 44 research research NOUN ajst-31990 43 1 [ [ X ajst-31990 43 2 8 8 NUM ajst-31990 43 3 ] ] PUNCT ajst-31990 43 4 . . PUNCT ajst-31990 44 1 the the DET ajst-31990 44 2 linear linear ADJ ajst-31990 44 3 part part NOUN ajst-31990 44 4 of of ADP ajst-31990 44 5 the the DET ajst-31990 44 6 logistic logistic ADJ ajst-31990 44 7 regression regression NOUN ajst-31990 44 8 is be AUX ajst-31990 44 9 𝑧 𝑧 PRON ajst-31990 44 10 = = X ajst-31990 44 11 𝑀0 𝑀0 NOUN ajst-31990 44 12 + + CCONJ ajst-31990 44 13 𝑀1π‘₯1 𝑀1π‘₯1 PROPN ajst-31990 44 14 + + PUNCT ajst-31990 44 15 𝑀2π‘₯2 𝑀2π‘₯2 X ajst-31990 44 16 + + CCONJ ajst-31990 44 17 β‹― β‹― VERB ajst-31990 44 18 + + VERB ajst-31990 44 19 𝑀𝑛π‘₯𝑛 𝑀𝑛π‘₯𝑛 NOUN ajst-31990 44 20 , , PUNCT ajst-31990 44 21 ( ( PUNCT ajst-31990 44 22 1 1 X ajst-31990 44 23 ) ) PUNCT ajst-31990 44 24 where where SCONJ ajst-31990 44 25 𝑀i 𝑀i ADV ajst-31990 44 26 is be AUX ajst-31990 44 27 the the DET ajst-31990 44 28 weight weight NOUN ajst-31990 44 29 of of ADP ajst-31990 44 30 the the DET ajst-31990 44 31 feature feature NOUN ajst-31990 44 32 π‘₯i π‘₯i ADP ajst-31990 44 33 , , PUNCT ajst-31990 44 34 the the DET ajst-31990 44 35 𝑀0 𝑀0 NOUN ajst-31990 44 36 is be AUX ajst-31990 44 37 the the DET ajst-31990 44 38 intercept intercept NOUN ajst-31990 44 39 term term NOUN ajst-31990 44 40 , , PUNCT ajst-31990 44 41 which which PRON ajst-31990 44 42 together together ADV ajst-31990 44 43 form form VERB ajst-31990 44 44 the the DET ajst-31990 44 45 hyperplane hyperplane NOUN ajst-31990 44 46 decision decision NOUN ajst-31990 44 47 boundary boundary NOUN ajst-31990 44 48 , , PUNCT ajst-31990 44 49 i.e. i.e. X ajst-31990 44 50 , , PUNCT ajst-31990 44 51 the the DET ajst-31990 44 52 classification classification NOUN ajst-31990 44 53 boundary boundary NOUN ajst-31990 44 54 is be AUX ajst-31990 44 55 defined define VERB ajst-31990 44 56 when when SCONJ ajst-31990 44 57 the the DET ajst-31990 44 58 linear linear ADJ ajst-31990 44 59 equation equation NOUN ajst-31990 44 60 𝑧 𝑧 NOUN ajst-31990 44 61 = = SYM ajst-31990 44 62 0 0 PUNCT ajst-31990 44 63 ( ( PUNCT ajst-31990 44 64 𝑀0 𝑀0 NOUN ajst-31990 44 65 + + NOUN ajst-31990 44 66 𝑀1π‘₯1 𝑀1π‘₯1 PROPN ajst-31990 44 67 + + CCONJ ajst-31990 44 68 β‹― β‹― VERB ajst-31990 44 69 + + CCONJ ajst-31990 44 70 𝑀𝑛π‘₯𝑛 𝑀𝑛π‘₯𝑛 NOUN ajst-31990 44 71 = = PUNCT ajst-31990 44 72 0 0 NUM ajst-31990 44 73 ) ) PUNCT ajst-31990 44 74 . . PUNCT ajst-31990 45 1 ( ( PUNCT ajst-31990 45 2 2 2 X ajst-31990 45 3 ) ) PUNCT ajst-31990 45 4 the the DET ajst-31990 45 5 key key NOUN ajst-31990 45 6 to to PART ajst-31990 45 7 probabilistic probabilistic VERB ajst-31990 45 8 mapping mapping NOUN ajst-31990 45 9 is be AUX ajst-31990 45 10 to to PART ajst-31990 45 11 map map VERB ajst-31990 45 12 the the DET ajst-31990 45 13 linear linear ADJ ajst-31990 45 14 output output NOUN ajst-31990 45 15 z z NOUN ajst-31990 45 16 to to ADP ajst-31990 45 17 the the DET ajst-31990 45 18 interval interval NOUN ajst-31990 46 1 [ [ X ajst-31990 46 2 0,1 0,1 X ajst-31990 46 3 ] ] PUNCT ajst-31990 46 4 using use VERB ajst-31990 46 5 a a DET ajst-31990 46 6 sigmoid sigmoid NOUN ajst-31990 46 7 function function NOUN ajst-31990 46 8 , , PUNCT ajst-31990 46 9 where where SCONJ ajst-31990 46 10 the the DET ajst-31990 46 11 sigmoid sigmoid NOUN ajst-31990 46 12 function function NOUN ajst-31990 46 13 is be AUX ajst-31990 46 14 𝑃(𝑦 𝑃(𝑦 PROPN ajst-31990 46 15 = = SYM ajst-31990 46 16 1 1 X ajst-31990 46 17 ) ) PUNCT ajst-31990 46 18 = = SYM ajst-31990 46 19 1 1 NUM ajst-31990 46 20 1+π‘’βˆ’π‘§ 1+π‘’βˆ’π‘§ NUM ajst-31990 46 21 , , PUNCT ajst-31990 46 22 ( ( PUNCT ajst-31990 46 23 3 3 X ajst-31990 46 24 ) ) PUNCT ajst-31990 46 25 which which PRON ajst-31990 46 26 maps map VERB ajst-31990 46 27 the the DET ajst-31990 46 28 linear linear ADJ ajst-31990 46 29 output output NOUN ajst-31990 46 30 z z NOUN ajst-31990 46 31 to to ADP ajst-31990 46 32 a a DET ajst-31990 46 33 probability probability NOUN ajst-31990 46 34 value value NOUN ajst-31990 46 35 , , PUNCT ajst-31990 46 36 with with ADP ajst-31990 46 37 a a DET ajst-31990 46 38 threshold threshold NOUN ajst-31990 46 39 of of ADP ajst-31990 46 40 0.5 0.5 NUM ajst-31990 46 41 determining determine VERB ajst-31990 46 42 the the DET ajst-31990 46 43 classification classification NOUN ajst-31990 46 44 result result NOUN ajst-31990 46 45 , , PUNCT ajst-31990 46 46 and and CCONJ ajst-31990 46 47 which which PRON ajst-31990 46 48 is be AUX ajst-31990 46 49 geometrically geometrically ADV ajst-31990 46 50 significant significant ADJ ajst-31990 46 51 to to PART ajst-31990 46 52 find find VERB ajst-31990 46 53 the the DET ajst-31990 46 54 hyperplane hyperplane NOUN ajst-31990 46 55 of of ADP ajst-31990 46 56 the the DET ajst-31990 46 57 feature feature NOUN ajst-31990 46 58 space space NOUN ajst-31990 46 59 that that PRON ajst-31990 46 60 distinguishes distinguish VERB ajst-31990 46 61 between between ADP ajst-31990 46 62 two two NUM ajst-31990 46 63 classes class NOUN ajst-31990 46 64 of of ADP ajst-31990 46 65 samples sample NOUN ajst-31990 46 66 . . PUNCT ajst-31990 47 1 2.3.2 2.3.2 NUM ajst-31990 47 2 xgboost xgboost X ajst-31990 47 3 gradient gradient ADJ ajst-31990 47 4 boosting boost VERB ajst-31990 47 5 tree tree NOUN ajst-31990 47 6 . . PUNCT ajst-31990 48 1 the the DET ajst-31990 48 2 xgboost xgboost PROPN ajst-31990 48 3 model model NOUN ajst-31990 48 4 is be AUX ajst-31990 48 5 one one NUM ajst-31990 48 6 of of ADP ajst-31990 48 7 the the DET ajst-31990 48 8 most most ADV ajst-31990 48 9 frequently frequently ADV ajst-31990 48 10 occurring occur VERB ajst-31990 48 11 models model NOUN ajst-31990 48 12 in in ADP ajst-31990 48 13 data datum NOUN ajst-31990 48 14 mining mining NOUN ajst-31990 48 15 , , PUNCT ajst-31990 48 16 where where SCONJ ajst-31990 48 17 multiple multiple ADJ ajst-31990 48 18 small small ADJ ajst-31990 48 19 trees tree NOUN ajst-31990 48 20 are be AUX ajst-31990 48 21 trained train VERB ajst-31990 48 22 by by ADP ajst-31990 48 23 iteratively iteratively ADV ajst-31990 48 24 adding add VERB ajst-31990 48 25 decision decision NOUN ajst-31990 48 26 trees tree NOUN ajst-31990 48 27 , , PUNCT ajst-31990 48 28 with with ADP ajst-31990 48 29 each each DET ajst-31990 48 30 tree tree NOUN ajst-31990 48 31 correcting correct VERB ajst-31990 48 32 the the DET ajst-31990 48 33 residuals residual NOUN ajst-31990 48 34 of of ADP ajst-31990 48 35 the the DET ajst-31990 48 36 previous previous ADJ ajst-31990 48 37 tree tree NOUN ajst-31990 48 38 , , PUNCT ajst-31990 48 39 and and CCONJ ajst-31990 48 40 the the DET ajst-31990 48 41 final final ADJ ajst-31990 48 42 prediction prediction NOUN ajst-31990 48 43 is be AUX ajst-31990 48 44 a a DET ajst-31990 48 45 weighted weighted ADJ ajst-31990 48 46 sum sum NOUN ajst-31990 48 47 of of ADP ajst-31990 48 48 all all DET ajst-31990 48 49 the the DET ajst-31990 48 50 trees tree NOUN ajst-31990 48 51 . . PUNCT ajst-31990 49 1 the the DET ajst-31990 49 2 objective objective ADJ ajst-31990 49 3 function function NOUN ajst-31990 49 4 is be AUX ajst-31990 49 5 total total ADJ ajst-31990 49 6 loss=βˆ‘loss loss=βˆ‘loss NUM ajst-31990 49 7 function+βˆ‘regularisation function+βˆ‘regularisation NOUN ajst-31990 49 8 term term NOUN ajst-31990 49 9 , , PUNCT ajst-31990 49 10 i.e. i.e. X ajst-31990 49 11 𝑂𝑏𝑗𝑒𝑐𝑑𝑖𝑣𝑒 𝑂𝑏𝑗𝑒𝑐𝑑𝑖𝑣𝑒 PROPN ajst-31990 49 12 = = SYM ajst-31990 49 13 βˆ‘ βˆ‘ PUNCT ajst-31990 49 14 𝐿(𝑦𝑖 𝐿(𝑦𝑖 PROPN ajst-31990 49 15 , , PUNCT ajst-31990 49 16 οΏ½ οΏ½ PROPN ajst-31990 49 17 Μ‚ Μ‚ VERB ajst-31990 49 18 οΏ½ οΏ½ NOUN ajst-31990 49 19 𝑖 𝑖 NUM ajst-31990 49 20 ) ) PUNCT ajst-31990 50 1 + + CCONJ ajst-31990 50 2 βˆ‘ βˆ‘ PROPN ajst-31990 50 3 𝛺(π‘“π‘˜)𝐾 𝛺(π‘“π‘˜)𝐾 ADJ ajst-31990 50 4 π‘˜=1 π‘˜=1 PROPN ajst-31990 50 5 𝑛 𝑛 PRON ajst-31990 50 6 𝑖=1 𝑖=1 PROPN ajst-31990 50 7 , , PUNCT ajst-31990 50 8 ( ( PUNCT ajst-31990 50 9 4 4 X ajst-31990 50 10 ) ) PUNCT ajst-31990 50 11 4 4 NUM ajst-31990 50 12 where where SCONJ ajst-31990 50 13 the the DET ajst-31990 50 14 loss loss NOUN ajst-31990 50 15 function function NOUN ajst-31990 50 16 l l NOUN ajst-31990 50 17 minimises minimise NOUN ajst-31990 50 18 the the DET ajst-31990 50 19 prediction prediction NOUN ajst-31990 50 20 error error NOUN ajst-31990 50 21 by by ADP ajst-31990 50 22 gradient gradient ADJ ajst-31990 50 23 descent descent NOUN ajst-31990 50 24 to to PART ajst-31990 50 25 ensure ensure VERB ajst-31990 50 26 that that SCONJ ajst-31990 50 27 the the DET ajst-31990 50 28 model model NOUN ajst-31990 50 29 prediction prediction NOUN ajst-31990 50 30 is be AUX ajst-31990 50 31 close close ADJ ajst-31990 50 32 to to ADP ajst-31990 50 33 the the DET ajst-31990 50 34 true true ADJ ajst-31990 50 35 value value NOUN ajst-31990 50 36 , , PUNCT ajst-31990 50 37 and and CCONJ ajst-31990 50 38 the the DET ajst-31990 50 39 regularisation regularisation NOUN ajst-31990 50 40 term term NOUN ajst-31990 50 41 Ο‰ Ο‰ PROPN ajst-31990 50 42 limits limit VERB ajst-31990 50 43 the the DET ajst-31990 50 44 number number NOUN ajst-31990 50 45 of of ADP ajst-31990 50 46 leaf leaf NOUN ajst-31990 50 47 nodes node NOUN ajst-31990 50 48 and and CCONJ ajst-31990 50 49 weight weight NOUN ajst-31990 50 50 values value NOUN ajst-31990 50 51 to to PART ajst-31990 50 52 control control VERB ajst-31990 50 53 the the DET ajst-31990 50 54 model model NOUN ajst-31990 50 55 complexity complexity NOUN ajst-31990 50 56 and and CCONJ ajst-31990 50 57 prevent prevent VERB ajst-31990 50 58 overfitting overfitting NOUN ajst-31990 50 59 . . PUNCT ajst-31990 51 1 the the DET ajst-31990 51 2 key key ADJ ajst-31990 51 3 optimisation optimisation NOUN ajst-31990 51 4 of of ADP ajst-31990 51 5 the the DET ajst-31990 51 6 xgboost xgboost PROPN ajst-31990 51 7 model model NOUN ajst-31990 51 8 is be AUX ajst-31990 51 9 the the DET ajst-31990 51 10 split split ADJ ajst-31990 51 11 gain gain NOUN ajst-31990 51 12 , , PUNCT ajst-31990 51 13 i.e. i.e. X ajst-31990 51 14 , , PUNCT ajst-31990 51 15 choosing choose VERB ajst-31990 51 16 the the DET ajst-31990 51 17 split split ADJ ajst-31990 51 18 point point NOUN ajst-31990 51 19 that that PRON ajst-31990 51 20 maximises maximise VERB ajst-31990 51 21 the the DET ajst-31990 51 22 gain gain NOUN ajst-31990 51 23 ( ( PUNCT ajst-31990 51 24 gain gain NOUN ajst-31990 51 25 ) ) PUNCT ajst-31990 51 26 , , PUNCT ajst-31990 51 27 allowing allow VERB ajst-31990 51 28 for for ADP ajst-31990 51 29 improved improved ADJ ajst-31990 51 30 prediction prediction NOUN ajst-31990 51 31 while while SCONJ ajst-31990 51 32 penalising penalise VERB ajst-31990 51 33 the the DET ajst-31990 51 34 complexity complexity NOUN ajst-31990 51 35 , , PUNCT ajst-31990 51 36 which which PRON ajst-31990 51 37 is be AUX ajst-31990 51 38 expressed express VERB ajst-31990 51 39 by by ADP ajst-31990 51 40 the the DET ajst-31990 51 41 formula formula NOUN ajst-31990 51 42 πΊπ‘Žπ‘–π‘› πΊπ‘Žπ‘–π‘› PROPN ajst-31990 51 43 = = PUNCT ajst-31990 51 44 left leave VERB ajst-31990 51 45 node node ADJ ajst-31990 51 46 score score NOUN ajst-31990 52 1 + + CCONJ ajst-31990 52 2 right right ADJ ajst-31990 52 3 node node ADJ ajst-31990 52 4 score score NOUN ajst-31990 52 5 parent parent NOUN ajst-31990 52 6 node node ADJ ajst-31990 52 7 score score NOUN ajst-31990 52 8 𝟐 𝟐 NUM ajst-31990 52 9 βˆ’Ξ³ βˆ’Ξ³ NOUN ajst-31990 52 10 ( ( PUNCT ajst-31990 52 11 5 5 NUM ajst-31990 52 12 ) ) PUNCT ajst-31990 52 13 the the DET ajst-31990 52 14 xgboost xgboost PROPN ajst-31990 52 15 model model PROPN ajst-31990 52 16 offers offer VERB ajst-31990 52 17 the the DET ajst-31990 52 18 advantages advantage NOUN ajst-31990 52 19 of of ADP ajst-31990 52 20 efficiency efficiency NOUN ajst-31990 52 21 , , PUNCT ajst-31990 52 22 flexibility flexibility NOUN ajst-31990 52 23 and and CCONJ ajst-31990 52 24 robustness robustness NOUN ajst-31990 52 25 , , PUNCT ajst-31990 52 26 can can AUX ajst-31990 52 27 be be AUX ajst-31990 52 28 presorted presorte VERB ajst-31990 52 29 with with ADP ajst-31990 52 30 chunked chunk VERB ajst-31990 52 31 parallel parallel ADJ ajst-31990 52 32 computation computation NOUN ajst-31990 52 33 , , PUNCT ajst-31990 52 34 supports support VERB ajst-31990 52 35 custom custom NOUN ajst-31990 52 36 loss loss NOUN ajst-31990 52 37 functions function NOUN ajst-31990 52 38 and and CCONJ ajst-31990 52 39 regularisation regularisation NOUN ajst-31990 52 40 , , PUNCT ajst-31990 52 41 and and CCONJ ajst-31990 52 42 automatically automatically ADV ajst-31990 52 43 handles handle VERB ajst-31990 52 44 missing miss VERB ajst-31990 52 45 values value NOUN ajst-31990 52 46 . . PUNCT ajst-31990 53 1 a a DET ajst-31990 53 2 comparative comparative ADJ ajst-31990 53 3 study study NOUN ajst-31990 53 4 by by ADP ajst-31990 53 5 santhanam santhanam PROPN ajst-31990 53 6 et et PROPN ajst-31990 53 7 al al PROPN ajst-31990 53 8 . . PROPN ajst-31990 53 9 indicated indicate VERB ajst-31990 53 10 that that SCONJ ajst-31990 53 11 xgboost xgboost PROPN ajst-31990 53 12 demonstrated demonstrate VERB ajst-31990 53 13 higher high ADJ ajst-31990 53 14 efficiency efficiency NOUN ajst-31990 53 15 and and CCONJ ajst-31990 53 16 accuracy accuracy NOUN ajst-31990 53 17 in in ADP ajst-31990 53 18 both both CCONJ ajst-31990 53 19 classification classification NOUN ajst-31990 53 20 and and CCONJ ajst-31990 53 21 regression regression NOUN ajst-31990 53 22 tasks task NOUN ajst-31990 53 23 , , PUNCT ajst-31990 53 24 with with ADP ajst-31990 53 25 significant significant ADJ ajst-31990 53 26 advantages advantage NOUN ajst-31990 53 27 especially especially ADV ajst-31990 53 28 when when SCONJ ajst-31990 53 29 dealing deal VERB ajst-31990 53 30 with with ADP ajst-31990 53 31 multivariate multivariate NOUN ajst-31990 53 32 features feature NOUN ajst-31990 53 33 , , PUNCT ajst-31990 53 34 providing provide VERB ajst-31990 53 35 an an DET ajst-31990 53 36 empirical empirical ADJ ajst-31990 53 37 reference reference NOUN ajst-31990 53 38 for for ADP ajst-31990 53 39 machine machine NOUN ajst-31990 53 40 learning learning NOUN ajst-31990 53 41 model model NOUN ajst-31990 53 42 selection selection NOUN ajst-31990 54 1 [ [ X ajst-31990 54 2 9 9 NUM ajst-31990 54 3 ] ] PUNCT ajst-31990 54 4 . . PUNCT ajst-31990 55 1 2.3.3 2.3.3 NUM ajst-31990 55 2 support support NOUN ajst-31990 55 3 vector vector NOUN ajst-31990 55 4 machines machine NOUN ajst-31990 55 5 ( ( PUNCT ajst-31990 55 6 svm svm PROPN ajst-31990 55 7 ) ) PUNCT ajst-31990 55 8 . . PUNCT ajst-31990 56 1 the the DET ajst-31990 56 2 support support NOUN ajst-31990 56 3 vector vector NOUN ajst-31990 56 4 machine machine NOUN ajst-31990 56 5 model model NOUN ajst-31990 56 6 maps map VERB ajst-31990 56 7 the the DET ajst-31990 56 8 data datum NOUN ajst-31990 56 9 to to ADP ajst-31990 56 10 a a DET ajst-31990 56 11 higher high ADJ ajst-31990 56 12 dimensional dimensional ADJ ajst-31990 56 13 space space NOUN ajst-31990 56 14 by by ADP ajst-31990 56 15 means mean NOUN ajst-31990 56 16 of of ADP ajst-31990 56 17 a a DET ajst-31990 56 18 kernel kernel NOUN ajst-31990 56 19 function function NOUN ajst-31990 56 20 that that PRON ajst-31990 56 21 searches search VERB ajst-31990 56 22 for for ADP ajst-31990 56 23 hyperplanes hyperplane NOUN ajst-31990 56 24 that that PRON ajst-31990 56 25 maximise maximise VERB ajst-31990 56 26 the the DET ajst-31990 56 27 margin margin NOUN ajst-31990 56 28 for for ADP ajst-31990 56 29 classification classification NOUN ajst-31990 56 30 , , PUNCT ajst-31990 56 31 where where SCONJ ajst-31990 56 32 the the DET ajst-31990 56 33 hyperplane hyperplane NOUN ajst-31990 56 34 equation equation NOUN ajst-31990 56 35 is be AUX ajst-31990 56 36 𝑓(π‘₯ 𝑓(π‘₯ NOUN ajst-31990 56 37 ) ) PUNCT ajst-31990 56 38 = = PUNCT ajst-31990 57 1 𝑀𝑇π‘₯ 𝑀𝑇π‘₯ X ajst-31990 57 2 + + NOUN ajst-31990 57 3 𝑏 𝑏 NOUN ajst-31990 57 4 = = SYM ajst-31990 57 5 0 0 PUNCT ajst-31990 57 6 ( ( PUNCT ajst-31990 57 7 w w NOUN ajst-31990 57 8 : : PUNCT ajst-31990 57 9 normal normal ADJ ajst-31990 57 10 vector vector NOUN ajst-31990 57 11 , , PUNCT ajst-31990 57 12 b b NOUN ajst-31990 57 13 : : PUNCT ajst-31990 57 14 bias bias NOUN ajst-31990 57 15 term term NOUN ajst-31990 57 16 ) ) PUNCT ajst-31990 57 17 , , PUNCT ajst-31990 57 18 ( ( PUNCT ajst-31990 57 19 6 6 NUM ajst-31990 57 20 ) ) PUNCT ajst-31990 57 21 and and CCONJ ajst-31990 57 22 the the DET ajst-31990 57 23 interval interval NOUN ajst-31990 57 24 equation equation NOUN ajst-31990 57 25 is be AUX ajst-31990 57 26 distance= distance= NUM ajst-31990 57 27 |𝑀𝑇π‘₯𝑖+𝑏| |𝑀𝑇π‘₯𝑖+𝑏| PUNCT ajst-31990 57 28 ‖𝑀‖ ‖𝑀‖ NOUN ajst-31990 57 29 , , PUNCT ajst-31990 57 30 ( ( PUNCT ajst-31990 57 31 7 7 X ajst-31990 57 32 ) ) PUNCT ajst-31990 57 33 the the DET ajst-31990 57 34 maximisation maximisation NOUN ajst-31990 57 35 of of ADP ajst-31990 57 36 the the DET ajst-31990 57 37 interval interval NOUN ajst-31990 57 38 is be AUX ajst-31990 57 39 achieved achieve VERB ajst-31990 57 40 by by ADP ajst-31990 57 41 maximising maximise VERB ajst-31990 57 42 the the DET ajst-31990 57 43 distance distance NOUN ajst-31990 57 44 of of ADP ajst-31990 57 45 the the DET ajst-31990 57 46 samples sample NOUN ajst-31990 57 47 to to ADP ajst-31990 57 48 the the DET ajst-31990 57 49 hyperplane hyperplane NOUN ajst-31990 57 50 , , PUNCT ajst-31990 57 51 and and CCONJ ajst-31990 57 52 the the DET ajst-31990 57 53 objective objective ADJ ajst-31990 57 54 function function NOUN ajst-31990 57 55 is be AUX ajst-31990 57 56 π‘šπ‘–π‘› π‘šπ‘–π‘› NOUN ajst-31990 57 57 𝑀,𝑏 𝑀,𝑏 NOUN ajst-31990 57 58 1 1 NUM ajst-31990 57 59 2 2 NUM ajst-31990 57 60 ‖𝑀‖2𝑠. ‖𝑀‖2𝑠. NOUN ajst-31990 57 61 𝑑. 𝑑. NOUN ajst-31990 57 62 , , PUNCT ajst-31990 57 63 ( ( PUNCT ajst-31990 57 64 8) 8) NUM ajst-31990 57 65 maximising maximise VERB ajst-31990 57 66 this this DET ajst-31990 57 67 value value NOUN ajst-31990 57 68 is be AUX ajst-31990 57 69 equivalent equivalent ADJ ajst-31990 57 70 to to ADP ajst-31990 57 71 minimising minimise VERB ajst-31990 57 72 the‖𝑀‖ the‖𝑀‖ ADP ajst-31990 57 73 , , PUNCT ajst-31990 57 74 with with ADP ajst-31990 57 75 the the DET ajst-31990 57 76 constraint constraint NOUN ajst-31990 57 77 𝑦𝑖(𝑀𝑇π‘₯𝑖 𝑦𝑖(𝑀𝑇π‘₯𝑖 NOUN ajst-31990 58 1 + + CCONJ ajst-31990 58 2 𝑏 𝑏 NOUN ajst-31990 58 3 ) ) PUNCT ajst-31990 58 4 β‰₯ β‰₯ NOUN ajst-31990 58 5 1(βˆ€π‘– 1(βˆ€π‘– NUM ajst-31990 58 6 ) ) PUNCT ajst-31990 58 7 , , PUNCT ajst-31990 58 8 ( ( PUNCT ajst-31990 58 9 9 9 X ajst-31990 58 10 ) ) PUNCT ajst-31990 58 11 ensuring ensure VERB ajst-31990 58 12 that that SCONJ ajst-31990 58 13 all all DET ajst-31990 58 14 samples sample NOUN ajst-31990 58 15 are be AUX ajst-31990 58 16 correctly correctly ADV ajst-31990 58 17 classified classified ADJ ajst-31990 58 18 and and CCONJ ajst-31990 58 19 that that SCONJ ajst-31990 58 20 the the DET ajst-31990 58 21 function function NOUN ajst-31990 58 22 interval interval NOUN ajst-31990 58 23 ( ( PUNCT ajst-31990 58 24 confidence confidence NOUN ajst-31990 58 25 level level NOUN ajst-31990 58 26 ) ) PUNCT ajst-31990 58 27 is be AUX ajst-31990 58 28 at at ADV ajst-31990 58 29 least least ADJ ajst-31990 58 30 1 1 NUM ajst-31990 58 31 . . PUNCT ajst-31990 59 1 the the DET ajst-31990 59 2 support support NOUN ajst-31990 59 3 vector vector NOUN ajst-31990 59 4 machine machine NOUN ajst-31990 59 5 model model NOUN ajst-31990 59 6 can can AUX ajst-31990 59 7 adapt adapt VERB ajst-31990 59 8 to to ADP ajst-31990 59 9 high high ADJ ajst-31990 59 10 - - PUNCT ajst-31990 59 11 dimensional dimensional ADJ ajst-31990 59 12 data datum NOUN ajst-31990 59 13 to to PART ajst-31990 59 14 avoid avoid VERB ajst-31990 59 15 overfitting overfitte VERB ajst-31990 59 16 , , PUNCT ajst-31990 59 17 and and CCONJ ajst-31990 59 18 the the DET ajst-31990 59 19 kernel kernel PROPN ajst-31990 59 20 function function NOUN ajst-31990 59 21 is be AUX ajst-31990 59 22 flexible flexible ADJ ajst-31990 59 23 to to PART ajst-31990 59 24 handle handle VERB ajst-31990 59 25 nonlinear nonlinear ADJ ajst-31990 59 26 problems problem NOUN ajst-31990 59 27 . . PUNCT ajst-31990 60 1 it it PRON ajst-31990 60 2 is be AUX ajst-31990 60 3 also also ADV ajst-31990 60 4 mentioned mention VERB ajst-31990 60 5 in in ADP ajst-31990 60 6 abdullah abdullah PROPN ajst-31990 60 7 et et PROPN ajst-31990 60 8 al al PROPN ajst-31990 60 9 . . PUNCT ajst-31990 61 1 that that DET ajst-31990 61 2 svm svm PROPN ajst-31990 61 3 performs perform VERB ajst-31990 61 4 well well ADV ajst-31990 61 5 in in ADP ajst-31990 61 6 small small ADJ ajst-31990 61 7 - - PUNCT ajst-31990 61 8 sample sample NOUN ajst-31990 61 9 , , PUNCT ajst-31990 61 10 high high ADJ ajst-31990 61 11 - - PUNCT ajst-31990 61 12 dimensional dimensional ADJ ajst-31990 61 13 and and CCONJ ajst-31990 61 14 non non ADJ ajst-31990 61 15 - - ADJ ajst-31990 61 16 linear linear ADJ ajst-31990 61 17 classification classification NOUN ajst-31990 61 18 tasks task NOUN ajst-31990 61 19 ( ( PUNCT ajst-31990 61 20 e.g. e.g. ADV ajst-31990 61 21 , , PUNCT ajst-31990 61 22 biomedical biomedical ADJ ajst-31990 61 23 , , PUNCT ajst-31990 61 24 image image NOUN ajst-31990 61 25 recognition recognition NOUN ajst-31990 61 26 ) ) PUNCT ajst-31990 61 27 , , PUNCT ajst-31990 61 28 but but CCONJ ajst-31990 61 29 its its PRON ajst-31990 61 30 efficiency efficiency NOUN ajst-31990 61 31 and and CCONJ ajst-31990 61 32 generalisability generalisability NOUN ajst-31990 61 33 are be AUX ajst-31990 61 34 constrained constrain VERB ajst-31990 61 35 by by ADP ajst-31990 61 36 data datum NOUN ajst-31990 61 37 and and CCONJ ajst-31990 61 38 parameters parameter NOUN ajst-31990 61 39 . . PUNCT ajst-31990 62 1 in in ADP ajst-31990 62 2 the the DET ajst-31990 62 3 future future NOUN ajst-31990 62 4 , , PUNCT ajst-31990 62 5 it it PRON ajst-31990 62 6 needs need VERB ajst-31990 62 7 to to PART ajst-31990 62 8 be be AUX ajst-31990 62 9 combined combine VERB ajst-31990 62 10 with with ADP ajst-31990 62 11 deep deep ADJ ajst-31990 62 12 learning learning NOUN ajst-31990 62 13 or or CCONJ ajst-31990 62 14 optimisation optimisation NOUN ajst-31990 62 15 algorithms algorithm NOUN ajst-31990 62 16 to to PART ajst-31990 62 17 improve improve VERB ajst-31990 62 18 the the DET ajst-31990 62 19 performance performance NOUN ajst-31990 62 20 of of ADP ajst-31990 62 21 large large ADJ ajst-31990 62 22 - - PUNCT ajst-31990 62 23 scale scale NOUN ajst-31990 62 24 applications application NOUN ajst-31990 62 25 [ [ X ajst-31990 62 26 10 10 NUM ajst-31990 62 27 ] ] PUNCT ajst-31990 62 28 . . PUNCT ajst-31990 63 1 2.4 2.4 NUM ajst-31990 63 2 . . PUNCT ajst-31990 64 1 data datum NOUN ajst-31990 64 2 sets set NOUN ajst-31990 64 3 and and CCONJ ajst-31990 64 4 feature feature NOUN ajst-31990 64 5 engineering engineer VERB ajst-31990 64 6 the the DET ajst-31990 64 7 dataset dataset NOUN ajst-31990 64 8 for for ADP ajst-31990 64 9 this this DET ajst-31990 64 10 study study NOUN ajst-31990 64 11 was be AUX ajst-31990 64 12 derived derive VERB ajst-31990 64 13 from from ADP ajst-31990 64 14 the the DET ajst-31990 64 15 uci uci PROPN ajst-31990 64 16 parkinson parkinson NOUN ajst-31990 64 17 's 's PART ajst-31990 64 18 speech speech NOUN ajst-31990 64 19 features feature VERB ajst-31990 64 20 dataset dataset NOUN ajst-31990 64 21 ( ( PUNCT ajst-31990 64 22 pd_speech_features.csv pd_speech_features.csv PROPN ajst-31990 64 23 ) ) PUNCT ajst-31990 64 24 , , PUNCT ajst-31990 64 25 which which PRON ajst-31990 64 26 contains contain VERB ajst-31990 64 27 754 754 NUM ajst-31990 64 28 - - PUNCT ajst-31990 64 29 dimensional dimensional ADJ ajst-31990 64 30 acoustic acoustic ADJ ajst-31990 64 31 features feature NOUN ajst-31990 64 32 ( ( PUNCT ajst-31990 64 33 fundamental fundamental ADJ ajst-31990 64 34 frequency frequency NOUN ajst-31990 64 35 perturbation perturbation NOUN ajst-31990 64 36 , , PUNCT ajst-31990 64 37 amplitude amplitude NOUN ajst-31990 64 38 perturbation perturbation NOUN ajst-31990 64 39 , , PUNCT ajst-31990 64 40 mfcc mfcc NOUN ajst-31990 64 41 , , PUNCT ajst-31990 64 42 etc etc X ajst-31990 64 43 . . X ajst-31990 64 44 ) ) PUNCT ajst-31990 65 1 and and CCONJ ajst-31990 65 2 nonlinear nonlinear ADJ ajst-31990 65 3 dynamical dynamical ADJ ajst-31990 65 4 features feature NOUN ajst-31990 65 5 . . PUNCT ajst-31990 66 1 its its PRON ajst-31990 66 2 data datum NOUN ajst-31990 66 3 is be AUX ajst-31990 66 4 characterised characterise VERB ajst-31990 66 5 by by ADP ajst-31990 66 6 high high ADJ ajst-31990 66 7 dimensionality dimensionality NOUN ajst-31990 66 8 and and CCONJ ajst-31990 66 9 category category NOUN ajst-31990 66 10 imbalance imbalance NOUN ajst-31990 66 11 , , PUNCT ajst-31990 66 12 and and CCONJ ajst-31990 66 13 the the DET ajst-31990 66 14 processing processing NOUN ajst-31990 66 15 goal goal NOUN ajst-31990 66 16 is be AUX ajst-31990 66 17 to to PART ajst-31990 66 18 build build VERB ajst-31990 66 19 a a DET ajst-31990 66 20 lightweight lightweight ADJ ajst-31990 66 21 classification classification NOUN ajst-31990 66 22 model model NOUN ajst-31990 66 23 ( ( PUNCT ajst-31990 66 24 from from ADP ajst-31990 66 25 754 754 NUM ajst-31990 66 26 dimensions dimension NOUN ajst-31990 66 27 down down ADV ajst-31990 66 28 to to ADP ajst-31990 66 29 30 30 NUM ajst-31990 66 30 dimensions dimension NOUN ajst-31990 66 31 ) ) PUNCT ajst-31990 66 32 for for ADP ajst-31990 66 33 automated automate VERB ajst-31990 66 34 screening screening NOUN ajst-31990 66 35 of of ADP ajst-31990 66 36 pd pd PROPN ajst-31990 66 37 by by ADP ajst-31990 66 38 means mean NOUN ajst-31990 66 39 of of ADP ajst-31990 66 40 speech speech NOUN ajst-31990 66 41 features feature NOUN ajst-31990 66 42 . . PUNCT ajst-31990 67 1 the the DET ajst-31990 67 2 feature feature NOUN ajst-31990 67 3 engineering engineering NOUN ajst-31990 67 4 of of ADP ajst-31990 67 5 the the DET ajst-31990 67 6 dataset dataset NOUN ajst-31990 67 7 is be AUX ajst-31990 67 8 divided divide VERB ajst-31990 67 9 into into ADP ajst-31990 67 10 three three NUM ajst-31990 67 11 steps step NOUN ajst-31990 67 12 , , PUNCT ajst-31990 67 13 the the DET ajst-31990 67 14 first first ADJ ajst-31990 67 15 is be AUX ajst-31990 67 16 data data NOUN ajst-31990 67 17 aggregation aggregation NOUN ajst-31990 67 18 , , PUNCT ajst-31990 67 19 grouping group VERB ajst-31990 67 20 patients patient NOUN ajst-31990 67 21 by by ADP ajst-31990 67 22 their their PRON ajst-31990 67 23 ids id NOUN ajst-31990 67 24 to to PART ajst-31990 67 25 take take VERB ajst-31990 67 26 the the DET ajst-31990 67 27 mean mean ADJ ajst-31990 67 28 value value NOUN ajst-31990 67 29 , , PUNCT ajst-31990 67 30 solving solve VERB ajst-31990 67 31 the the DET ajst-31990 67 32 problem problem NOUN ajst-31990 67 33 of of ADP ajst-31990 67 34 multiple multiple ADJ ajst-31990 67 35 sampling sampling NOUN ajst-31990 67 36 of of ADP ajst-31990 67 37 the the DET ajst-31990 67 38 same same ADJ ajst-31990 67 39 patient patient NOUN ajst-31990 67 40 and and CCONJ ajst-31990 67 41 avoiding avoid VERB ajst-31990 67 42 data datum NOUN ajst-31990 67 43 leakage leakage NOUN ajst-31990 67 44 . . PUNCT ajst-31990 68 1 next next ADJ ajst-31990 68 2 is be AUX ajst-31990 68 3 the the DET ajst-31990 68 4 high high ADJ ajst-31990 68 5 correlation correlation NOUN ajst-31990 68 6 feature feature NOUN ajst-31990 68 7 filtering filtering NOUN ajst-31990 68 8 , , PUNCT ajst-31990 68 9 which which PRON ajst-31990 68 10 calculates calculate VERB ajst-31990 68 11 5 5 NUM ajst-31990 68 12 the the DET ajst-31990 68 13 pearson pearson NOUN ajst-31990 68 14 correlation correlation NOUN ajst-31990 68 15 coefficient coefficient NOUN ajst-31990 68 16 between between ADP ajst-31990 68 17 features feature NOUN ajst-31990 68 18 through through ADP ajst-31990 68 19 double double ADJ ajst-31990 68 20 loops loop NOUN ajst-31990 68 21 , , PUNCT ajst-31990 68 22 dynamically dynamically ADV ajst-31990 68 23 removes remove VERB ajst-31990 68 24 redundant redundant ADJ ajst-31990 68 25 features feature NOUN ajst-31990 68 26 with with ADP ajst-31990 68 27 correlation correlation NOUN ajst-31990 68 28 coefficients coefficient NOUN ajst-31990 68 29 > > X ajst-31990 68 30 0.7 0.7 NUM ajst-31990 68 31 , , PUNCT ajst-31990 68 32 and and CCONJ ajst-31990 68 33 retains retain VERB ajst-31990 68 34 120 120 NUM ajst-31990 68 35 features feature NOUN ajst-31990 68 36 to to PART ajst-31990 68 37 reduce reduce VERB ajst-31990 68 38 the the DET ajst-31990 68 39 computational computational ADJ ajst-31990 68 40 cost cost NOUN ajst-31990 68 41 and and CCONJ ajst-31990 68 42 improve improve VERB ajst-31990 68 43 the the DET ajst-31990 68 44 generalisation generalisation NOUN ajst-31990 68 45 ability ability NOUN ajst-31990 68 46 of of ADP ajst-31990 68 47 the the DET ajst-31990 68 48 model model NOUN ajst-31990 68 49 . . PUNCT ajst-31990 69 1 finally finally ADV ajst-31990 69 2 , , PUNCT ajst-31990 69 3 normalisation normalisation NOUN ajst-31990 69 4 and and CCONJ ajst-31990 69 5 feature feature NOUN ajst-31990 69 6 selection selection NOUN ajst-31990 69 7 , , PUNCT ajst-31990 69 8 using use VERB ajst-31990 69 9 minmaxscaler minmaxscaler ADJ ajst-31990 69 10 normalisation normalisation NOUN ajst-31990 69 11 to to PART ajst-31990 69 12 scale scale VERB ajst-31990 69 13 the the DET ajst-31990 69 14 feature feature NOUN ajst-31990 69 15 values value NOUN ajst-31990 69 16 to to ADP ajst-31990 69 17 [ [ X ajst-31990 69 18 0,1 0,1 NUM ajst-31990 69 19 ] ] PUNCT ajst-31990 69 20 to to PART ajst-31990 69 21 avoid avoid VERB ajst-31990 69 22 large large ADJ ajst-31990 69 23 - - PUNCT ajst-31990 69 24 valued value VERB ajst-31990 69 25 features feature NOUN ajst-31990 69 26 dominating dominate VERB ajst-31990 69 27 the the DET ajst-31990 69 28 test test NOUN ajst-31990 69 29 , , PUNCT ajst-31990 69 30 and and CCONJ ajst-31990 69 31 then then ADV ajst-31990 69 32 using use VERB ajst-31990 69 33 selectkbest(chi2 selectkbest(chi2 NOUN ajst-31990 69 34 , , PUNCT ajst-31990 69 35 k=30 k=30 PROPN ajst-31990 69 36 ) ) PUNCT ajst-31990 69 37 to to PART ajst-31990 69 38 select select VERB ajst-31990 69 39 the the DET ajst-31990 69 40 30 30 NUM ajst-31990 69 41 features feature NOUN ajst-31990 69 42 with with ADP ajst-31990 69 43 the the DET ajst-31990 69 44 highest high ADJ ajst-31990 69 45 chi chi ADJ ajst-31990 69 46 - - PUNCT ajst-31990 69 47 square square ADJ ajst-31990 69 48 test test NOUN ajst-31990 69 49 scores score NOUN ajst-31990 69 50 , , PUNCT ajst-31990 69 51 retaining retain VERB ajst-31990 69 52 the the DET ajst-31990 69 53 features feature NOUN ajst-31990 69 54 that that PRON ajst-31990 69 55 are be AUX ajst-31990 69 56 most most ADV ajst-31990 69 57 relevant relevant ADJ ajst-31990 69 58 to to ADP ajst-31990 69 59 the the DET ajst-31990 69 60 category category NOUN ajst-31990 69 61 . . PUNCT ajst-31990 70 1 2.5 2.5 NUM ajst-31990 70 2 . . PUNCT ajst-31990 71 1 training training NOUN ajst-31990 71 2 models model NOUN ajst-31990 71 3 in in ADP ajst-31990 71 4 this this DET ajst-31990 71 5 study study NOUN ajst-31990 71 6 , , PUNCT ajst-31990 71 7 the the DET ajst-31990 71 8 dataset dataset NOUN ajst-31990 71 9 is be AUX ajst-31990 71 10 divided divide VERB ajst-31990 71 11 into into ADP ajst-31990 71 12 training training NOUN ajst-31990 71 13 and and CCONJ ajst-31990 71 14 validation validation NOUN ajst-31990 71 15 sets set NOUN ajst-31990 71 16 using use VERB ajst-31990 71 17 an an DET ajst-31990 71 18 8:2 8:2 NUM ajst-31990 71 19 ratio ratio NOUN ajst-31990 71 20 , , PUNCT ajst-31990 71 21 where where SCONJ ajst-31990 71 22 the the DET ajst-31990 71 23 training training NOUN ajst-31990 71 24 set set NOUN ajst-31990 71 25 is be AUX ajst-31990 71 26 used use VERB ajst-31990 71 27 for for ADP ajst-31990 71 28 model model NOUN ajst-31990 71 29 learning learning NOUN ajst-31990 71 30 and and CCONJ ajst-31990 71 31 the the DET ajst-31990 71 32 validation validation NOUN ajst-31990 71 33 set set NOUN ajst-31990 71 34 is be AUX ajst-31990 71 35 used use VERB ajst-31990 71 36 for for ADP ajst-31990 71 37 performance performance NOUN ajst-31990 71 38 evaluation evaluation NOUN ajst-31990 71 39 . . PUNCT ajst-31990 72 1 to to PART ajst-31990 72 2 address address VERB ajst-31990 72 3 the the DET ajst-31990 72 4 category category NOUN ajst-31990 72 5 imbalance imbalance NOUN ajst-31990 72 6 problem problem NOUN ajst-31990 72 7 , , PUNCT ajst-31990 72 8 the the DET ajst-31990 72 9 minority minority NOUN ajst-31990 72 10 class class NOUN ajst-31990 72 11 samples sample NOUN ajst-31990 72 12 ( ( PUNCT ajst-31990 72 13 i.e. i.e. X ajst-31990 72 14 , , PUNCT ajst-31990 72 15 diseased diseased ADJ ajst-31990 72 16 samples sample NOUN ajst-31990 72 17 ) ) PUNCT ajst-31990 72 18 of of ADP ajst-31990 72 19 the the DET ajst-31990 72 20 training training NOUN ajst-31990 72 21 set set NOUN ajst-31990 72 22 were be AUX ajst-31990 72 23 replicated replicate VERB ajst-31990 72 24 using use VERB ajst-31990 72 25 randomoversampler randomoversampler NOUN ajst-31990 72 26 to to PART ajst-31990 72 27 achieve achieve VERB ajst-31990 72 28 a a DET ajst-31990 72 29 1:1 1:1 NUM ajst-31990 72 30 ratio ratio NOUN ajst-31990 72 31 of of ADP ajst-31990 72 32 diseased diseased ADJ ajst-31990 72 33 to to ADP ajst-31990 72 34 healthy healthy ADJ ajst-31990 72 35 samples sample NOUN ajst-31990 72 36 in in ADP ajst-31990 72 37 the the DET ajst-31990 72 38 training training NOUN ajst-31990 72 39 set set NOUN ajst-31990 72 40 . . PUNCT ajst-31990 73 1 the the DET ajst-31990 73 2 validation validation NOUN ajst-31990 73 3 set set NOUN ajst-31990 73 4 maintained maintain VERB ajst-31990 73 5 the the DET ajst-31990 73 6 original original ADJ ajst-31990 73 7 data datum NOUN ajst-31990 73 8 distribution distribution NOUN ajst-31990 73 9 to to PART ajst-31990 73 10 avoid avoid VERB ajst-31990 73 11 validation validation NOUN ajst-31990 73 12 bias bias NOUN ajst-31990 73 13 introduced introduce VERB ajst-31990 73 14 by by ADP ajst-31990 73 15 oversampling oversample VERB ajst-31990 73 16 . . PUNCT ajst-31990 74 1 for for ADP ajst-31990 74 2 model model NOUN ajst-31990 74 3 training training NOUN ajst-31990 74 4 and and CCONJ ajst-31990 74 5 evaluation evaluation NOUN ajst-31990 74 6 , , PUNCT ajst-31990 74 7 three three NUM ajst-31990 74 8 classical classical ADJ ajst-31990 74 9 classification classification NOUN ajst-31990 74 10 models model NOUN ajst-31990 74 11 , , PUNCT ajst-31990 74 12 logistic logistic ADJ ajst-31990 74 13 regression regression NOUN ajst-31990 74 14 , , PUNCT ajst-31990 74 15 xgboost xgboost ADV ajst-31990 74 16 and and CCONJ ajst-31990 74 17 svm svm ADJ ajst-31990 74 18 , , PUNCT ajst-31990 74 19 were be AUX ajst-31990 74 20 used use VERB ajst-31990 74 21 for for ADP ajst-31990 74 22 training training NOUN ajst-31990 74 23 respectively respectively ADV ajst-31990 74 24 . . PUNCT ajst-31990 75 1 each each DET ajst-31990 75 2 model model NOUN ajst-31990 75 3 is be AUX ajst-31990 75 4 parameter parameter NOUN ajst-31990 75 5 optimised optimise VERB ajst-31990 75 6 based base VERB ajst-31990 75 7 on on ADP ajst-31990 75 8 the the DET ajst-31990 75 9 training training NOUN ajst-31990 75 10 set set NOUN ajst-31990 75 11 , , PUNCT ajst-31990 75 12 and and CCONJ ajst-31990 75 13 the the DET ajst-31990 75 14 classification classification NOUN ajst-31990 75 15 performance performance NOUN ajst-31990 75 16 ( ( PUNCT ajst-31990 75 17 e.g. e.g. ADV ajst-31990 75 18 , , PUNCT ajst-31990 75 19 precision precision NOUN ajst-31990 75 20 , , PUNCT ajst-31990 75 21 recall recall NOUN ajst-31990 75 22 , , PUNCT ajst-31990 75 23 etc etc X ajst-31990 75 24 . . X ajst-31990 75 25 ) ) PUNCT ajst-31990 75 26 is be AUX ajst-31990 75 27 evaluated evaluate VERB ajst-31990 75 28 on on ADP ajst-31990 75 29 the the DET ajst-31990 75 30 validation validation NOUN ajst-31990 75 31 set set NOUN ajst-31990 75 32 , , PUNCT ajst-31990 75 33 and and CCONJ ajst-31990 75 34 the the DET ajst-31990 75 35 generalisation generalisation NOUN ajst-31990 75 36 ability ability NOUN ajst-31990 75 37 is be AUX ajst-31990 75 38 further far ADV ajst-31990 75 39 improved improve VERB ajst-31990 75 40 through through ADP ajst-31990 75 41 cross cross ADJ ajst-31990 75 42 - - ADJ ajst-31990 75 43 validation validation ADJ ajst-31990 75 44 and and CCONJ ajst-31990 75 45 hyperparameter hyperparameter NOUN ajst-31990 75 46 tuning tuning NOUN ajst-31990 75 47 . . PUNCT ajst-31990 76 1 this this DET ajst-31990 76 2 process process NOUN ajst-31990 76 3 balances balance VERB ajst-31990 76 4 data datum NOUN ajst-31990 76 5 balance balance NOUN ajst-31990 76 6 , , PUNCT ajst-31990 76 7 model model NOUN ajst-31990 76 8 diversity diversity NOUN ajst-31990 76 9 and and CCONJ ajst-31990 76 10 assessment assessment NOUN ajst-31990 76 11 rigour rigour NOUN ajst-31990 76 12 , , PUNCT ajst-31990 76 13 laying lay VERB ajst-31990 76 14 the the DET ajst-31990 76 15 foundation foundation NOUN ajst-31990 76 16 for for ADP ajst-31990 76 17 subsequent subsequent ADJ ajst-31990 76 18 model model NOUN ajst-31990 76 19 selection selection NOUN ajst-31990 76 20 and and CCONJ ajst-31990 76 21 optimisation optimisation NOUN ajst-31990 76 22 . . PUNCT ajst-31990 77 1 2.6 2.6 NUM ajst-31990 77 2 . . PUNCT ajst-31990 77 3 assessment assessment NOUN ajst-31990 77 4 of of ADP ajst-31990 77 5 indicators indicator NOUN ajst-31990 77 6 2.6.1 2.6.1 NUM ajst-31990 77 7 roc roc NOUN ajst-31990 77 8 - - PUNCT ajst-31990 77 9 auc auc NOUN ajst-31990 77 10 . . PUNCT ajst-31990 78 1 in in ADP ajst-31990 78 2 this this DET ajst-31990 78 3 study study NOUN ajst-31990 78 4 , , PUNCT ajst-31990 78 5 roc roc NOUN ajst-31990 78 6 - - PUNCT ajst-31990 78 7 auc auc NOUN ajst-31990 78 8 ( ( PUNCT ajst-31990 78 9 receiver receiver NOUN ajst-31990 78 10 operating operate VERB ajst-31990 78 11 characteristic characteristic ADJ ajst-31990 78 12 area area NOUN ajst-31990 78 13 under under ADP ajst-31990 78 14 the the DET ajst-31990 78 15 curve curve NOUN ajst-31990 78 16 ) ) PUNCT ajst-31990 78 17 is be AUX ajst-31990 78 18 used use VERB ajst-31990 78 19 as as ADP ajst-31990 78 20 the the DET ajst-31990 78 21 core core NOUN ajst-31990 78 22 index index NOUN ajst-31990 78 23 for for ADP ajst-31990 78 24 classification classification NOUN ajst-31990 78 25 model model NOUN ajst-31990 78 26 performance performance NOUN ajst-31990 78 27 evaluation evaluation NOUN ajst-31990 78 28 , , PUNCT ajst-31990 78 29 and and CCONJ ajst-31990 78 30 the the DET ajst-31990 78 31 model model NOUN ajst-31990 78 32 performance performance NOUN ajst-31990 78 33 is be AUX ajst-31990 78 34 comprehensively comprehensively ADV ajst-31990 78 35 evaluated evaluate VERB ajst-31990 78 36 by by ADP ajst-31990 78 37 dynamically dynamically ADV ajst-31990 78 38 adjusting adjust VERB ajst-31990 78 39 the the DET ajst-31990 78 40 classification classification NOUN ajst-31990 78 41 threshold threshold NOUN ajst-31990 78 42 . . PUNCT ajst-31990 79 1 the the DET ajst-31990 79 2 horizontal horizontal ADJ ajst-31990 79 3 axis axis NOUN ajst-31990 79 4 of of ADP ajst-31990 79 5 the the DET ajst-31990 79 6 roc roc PROPN ajst-31990 79 7 curve curve NOUN ajst-31990 79 8 ( ( PUNCT ajst-31990 79 9 fpr fpr NOUN ajst-31990 79 10 / / SYM ajst-31990 79 11 false false ADJ ajst-31990 79 12 positive positive ADJ ajst-31990 79 13 rate rate NOUN ajst-31990 79 14 ) ) PUNCT ajst-31990 79 15 represents represent VERB ajst-31990 79 16 the the DET ajst-31990 79 17 false false ADJ ajst-31990 79 18 positive positive ADJ ajst-31990 79 19 rate rate NOUN ajst-31990 79 20 , , PUNCT ajst-31990 79 21 i.e. i.e. X ajst-31990 79 22 , , PUNCT ajst-31990 79 23 the the DET ajst-31990 79 24 probability probability NOUN ajst-31990 79 25 that that SCONJ ajst-31990 79 26 a a DET ajst-31990 79 27 negative negative ADJ ajst-31990 79 28 case case NOUN ajst-31990 79 29 will will AUX ajst-31990 79 30 be be AUX ajst-31990 79 31 misclassified misclassifie VERB ajst-31990 79 32 as as ADP ajst-31990 79 33 a a DET ajst-31990 79 34 positive positive ADJ ajst-31990 79 35 case case NOUN ajst-31990 79 36 , , PUNCT ajst-31990 79 37 with with ADP ajst-31990 79 38 the the DET ajst-31990 79 39 formula formula NOUN ajst-31990 79 40 𝑇𝑃𝑅 𝑇𝑃𝑅 NOUN ajst-31990 79 41 = = SYM ajst-31990 79 42 𝑇𝑃 𝑇𝑃 PROPN ajst-31990 79 43 𝑇𝑃+𝐹𝑁 𝑇𝑃+𝐹𝑁 NOUN ajst-31990 79 44 ( ( PUNCT ajst-31990 79 45 tp tp NOUN ajst-31990 79 46 , , PUNCT ajst-31990 79 47 true true ADJ ajst-31990 79 48 positive positive ADJ ajst-31990 79 49 ) ) PUNCT ajst-31990 79 50 ( ( PUNCT ajst-31990 79 51 fn fn NOUN ajst-31990 79 52 , , PUNCT ajst-31990 79 53 false false ADJ ajst-31990 79 54 negative negative NOUN ajst-31990 79 55 ) ) PUNCT ajst-31990 79 56 ; ; PUNCT ajst-31990 79 57 ( ( PUNCT ajst-31990 79 58 10 10 NUM ajst-31990 79 59 ) ) PUNCT ajst-31990 79 60 the the DET ajst-31990 79 61 vertical vertical ADJ ajst-31990 79 62 axis axis NOUN ajst-31990 79 63 ( ( PUNCT ajst-31990 79 64 tpr tpr NOUN ajst-31990 79 65 / / SYM ajst-31990 79 66 truepositive truepositive ADJ ajst-31990 79 67 rate rate NOUN ajst-31990 79 68 ) ) PUNCT ajst-31990 79 69 represents represent VERB ajst-31990 79 70 the the DET ajst-31990 79 71 true true ADJ ajst-31990 79 72 positive positive ADJ ajst-31990 79 73 rate rate NOUN ajst-31990 79 74 , , PUNCT ajst-31990 79 75 i.e. i.e. X ajst-31990 79 76 , , PUNCT ajst-31990 79 77 the the DET ajst-31990 79 78 probability probability NOUN ajst-31990 79 79 that that SCONJ ajst-31990 79 80 a a DET ajst-31990 79 81 positive positive ADJ ajst-31990 79 82 case case NOUN ajst-31990 79 83 is be AUX ajst-31990 79 84 correctly correctly ADV ajst-31990 79 85 identified identify VERB ajst-31990 79 86 , , PUNCT ajst-31990 79 87 with with ADP ajst-31990 79 88 the the DET ajst-31990 79 89 formula formula NOUN ajst-31990 79 90 𝐹𝑃𝑅 𝐹𝑃𝑅 NOUN ajst-31990 79 91 = = SYM ajst-31990 79 92 𝐹𝑃 𝐹𝑃 PROPN ajst-31990 79 93 𝐹𝑃+𝑇𝑁 𝐹𝑃+𝑇𝑁 X ajst-31990 79 94 ( ( PUNCT ajst-31990 79 95 fp fp INTJ ajst-31990 79 96 , , PUNCT ajst-31990 79 97 false false ADJ ajst-31990 79 98 positive positive ADJ ajst-31990 79 99 ) ) PUNCT ajst-31990 79 100 ( ( PUNCT ajst-31990 79 101 tn tn NOUN ajst-31990 79 102 , , PUNCT ajst-31990 79 103 true true ADJ ajst-31990 79 104 negative negative NOUN ajst-31990 79 105 ) ) PUNCT ajst-31990 79 106 . . PUNCT ajst-31990 80 1 ( ( PUNCT ajst-31990 80 2 11 11 NUM ajst-31990 80 3 ) ) PUNCT ajst-31990 80 4 each each DET ajst-31990 80 5 point point NOUN ajst-31990 80 6 on on ADP ajst-31990 80 7 the the DET ajst-31990 80 8 curve curve NOUN ajst-31990 80 9 corresponds correspond VERB ajst-31990 80 10 to to ADP ajst-31990 80 11 a a DET ajst-31990 80 12 threshold threshold NOUN ajst-31990 80 13 , , PUNCT ajst-31990 80 14 and and CCONJ ajst-31990 80 15 ideally ideally ADV ajst-31990 80 16 the the DET ajst-31990 80 17 curve curve NOUN ajst-31990 80 18 converges converge VERB ajst-31990 80 19 to to ADP ajst-31990 80 20 the the DET ajst-31990 80 21 upper upper ADJ ajst-31990 80 22 left left ADJ ajst-31990 80 23 corner corner NOUN ajst-31990 80 24 ( ( PUNCT ajst-31990 80 25 fpr fpr NOUN ajst-31990 80 26 = = SYM ajst-31990 80 27 0 0 NUM ajst-31990 80 28 and and CCONJ ajst-31990 80 29 tpr tpr PROPN ajst-31990 80 30 = = NOUN ajst-31990 80 31 1 1 NUM ajst-31990 80 32 ) ) PUNCT ajst-31990 80 33 , , PUNCT ajst-31990 80 34 reflecting reflect VERB ajst-31990 80 35 the the DET ajst-31990 80 36 model model NOUN ajst-31990 80 37 's 's PART ajst-31990 80 38 ability ability NOUN ajst-31990 80 39 to to PART ajst-31990 80 40 perfectly perfectly ADV ajst-31990 80 41 discriminate discriminate VERB ajst-31990 80 42 between between ADP ajst-31990 80 43 positive positive ADJ ajst-31990 80 44 and and CCONJ ajst-31990 80 45 negative negative ADJ ajst-31990 80 46 samples sample NOUN ajst-31990 80 47 . . PUNCT ajst-31990 81 1 the the DET ajst-31990 81 2 overall overall ADJ ajst-31990 81 3 classification classification NOUN ajst-31990 81 4 performance performance NOUN ajst-31990 81 5 of of ADP ajst-31990 81 6 the the DET ajst-31990 81 7 model model NOUN ajst-31990 81 8 under under ADP ajst-31990 81 9 all all DET ajst-31990 81 10 thresholds threshold NOUN ajst-31990 81 11 is be AUX ajst-31990 81 12 usually usually ADV ajst-31990 81 13 quantified quantify VERB ajst-31990 81 14 by by ADP ajst-31990 81 15 calculating calculate VERB ajst-31990 81 16 the the DET ajst-31990 81 17 area area NOUN ajst-31990 81 18 auc auc NOUN ajst-31990 81 19 under under ADP ajst-31990 81 20 roc roc PROPN ajst-31990 81 21 to to PART ajst-31990 81 22 evaluate evaluate VERB ajst-31990 81 23 the the DET ajst-31990 81 24 model model NOUN ajst-31990 81 25 's 's PART ajst-31990 81 26 merit merit NOUN ajst-31990 81 27 . . PUNCT ajst-31990 82 1 the the DET ajst-31990 82 2 best good ADJ ajst-31990 82 3 classifier classifier NOUN ajst-31990 82 4 performance performance NOUN ajst-31990 82 5 is be AUX ajst-31990 82 6 achieved achieve VERB ajst-31990 82 7 when when SCONJ ajst-31990 82 8 the the DET ajst-31990 82 9 auc auc NOUN ajst-31990 82 10 is be AUX ajst-31990 82 11 1.the 1.the DET ajst-31990 82 12 advantage advantage NOUN ajst-31990 82 13 is be AUX ajst-31990 82 14 that that SCONJ ajst-31990 82 15 it it PRON ajst-31990 82 16 combines combine VERB ajst-31990 82 17 all all DET ajst-31990 82 18 threshold threshold NOUN ajst-31990 82 19 performances performance NOUN ajst-31990 82 20 and and CCONJ ajst-31990 82 21 is be AUX ajst-31990 82 22 not not PART ajst-31990 82 23 affected affect VERB ajst-31990 82 24 by by ADP ajst-31990 82 25 category category NOUN ajst-31990 82 26 imbalance imbalance NOUN ajst-31990 82 27 . . PUNCT ajst-31990 83 1 compared compare VERB ajst-31990 83 2 to to ADP ajst-31990 83 3 accuracy accuracy NOUN ajst-31990 83 4 , , PUNCT ajst-31990 83 5 roc roc NOUN ajst-31990 83 6 - - PUNCT ajst-31990 83 7 auc auc NOUN ajst-31990 83 8 is be AUX ajst-31990 83 9 more more ADV ajst-31990 83 10 robust robust ADJ ajst-31990 83 11 in in ADP ajst-31990 83 12 such such ADJ ajst-31990 83 13 sample sample NOUN ajst-31990 83 14 category category NOUN ajst-31990 83 15 imbalance imbalance NOUN ajst-31990 83 16 scenarios scenario NOUN ajst-31990 83 17 such such ADJ ajst-31990 83 18 as as ADP ajst-31990 83 19 medical medical ADJ ajst-31990 83 20 diagnosis diagnosis NOUN ajst-31990 83 21 , , PUNCT ajst-31990 83 22 objectively objectively ADV ajst-31990 83 23 assesses assess VERB ajst-31990 83 24 the the DET ajst-31990 83 25 model model NOUN ajst-31990 83 26 's 's PART ajst-31990 83 27 ability ability NOUN ajst-31990 83 28 to to PART ajst-31990 83 29 recognise recognise VERB ajst-31990 83 30 a a DET ajst-31990 83 31 small small ADJ ajst-31990 83 32 number number NOUN ajst-31990 83 33 of of ADP ajst-31990 83 34 classes class NOUN ajst-31990 83 35 , , PUNCT ajst-31990 83 36 and and CCONJ ajst-31990 83 37 provides provide VERB ajst-31990 83 38 a a DET ajst-31990 83 39 uniform uniform ADJ ajst-31990 83 40 standard standard NOUN ajst-31990 83 41 for for ADP ajst-31990 83 42 multi multi ADJ ajst-31990 83 43 - - ADJ ajst-31990 83 44 model model ADJ ajst-31990 83 45 comparison comparison NOUN ajst-31990 83 46 . . PUNCT ajst-31990 84 1 2.6.2 2.6.2 NUM ajst-31990 84 2 confusion confusion NOUN ajst-31990 84 3 matrix matrix NOUN ajst-31990 84 4 . . PUNCT ajst-31990 85 1 confusion confusion NOUN ajst-31990 85 2 matri matri PROPN ajst-31990 85 3 is be AUX ajst-31990 85 4 a a DET ajst-31990 85 5 core core NOUN ajst-31990 85 6 tool tool NOUN ajst-31990 85 7 for for ADP ajst-31990 85 8 classification classification NOUN ajst-31990 85 9 model model NOUN ajst-31990 85 10 performance performance NOUN ajst-31990 85 11 evaluation evaluation NOUN ajst-31990 85 12 , , PUNCT ajst-31990 85 13 by by ADP ajst-31990 85 14 comparing compare VERB ajst-31990 85 15 model model NOUN ajst-31990 85 16 predictions prediction NOUN ajst-31990 85 17 with with ADP ajst-31990 85 18 real real ADJ ajst-31990 85 19 labels label NOUN ajst-31990 85 20 to to PART ajst-31990 85 21 form form VERB ajst-31990 85 22 a a DET ajst-31990 85 23 two two NUM ajst-31990 85 24 - - PUNCT ajst-31990 85 25 dimensional dimensional ADJ ajst-31990 85 26 table table NOUN ajst-31990 85 27 visualising visualise VERB ajst-31990 85 28 classification classification NOUN ajst-31990 85 29 performance performance NOUN ajst-31990 85 30 . . PUNCT ajst-31990 86 1 in in ADP ajst-31990 86 2 the the DET ajst-31990 86 3 binary binary ADJ ajst-31990 86 4 classification classification NOUN ajst-31990 86 5 problem problem NOUN ajst-31990 86 6 , , PUNCT ajst-31990 86 7 its its PRON ajst-31990 86 8 standard standard ADJ ajst-31990 86 9 form form NOUN ajst-31990 86 10 contains contain VERB ajst-31990 86 11 four four NUM ajst-31990 86 12 key key ADJ ajst-31990 86 13 metrics metric NOUN ajst-31990 86 14 : : PUNCT ajst-31990 86 15 true true ADJ ajst-31990 86 16 positive positive ADJ ajst-31990 86 17 ( ( PUNCT ajst-31990 86 18 tp tp NOUN ajst-31990 86 19 ) ) PUNCT ajst-31990 86 20 denotes denote NOUN ajst-31990 86 21 correctly correctly ADV ajst-31990 86 22 identified identify VERB ajst-31990 86 23 patient patient NOUN ajst-31990 86 24 samples sample NOUN ajst-31990 86 25 , , PUNCT ajst-31990 86 26 which which PRON ajst-31990 86 27 need need VERB ajst-31990 86 28 to to PART ajst-31990 86 29 be be AUX ajst-31990 86 30 maximised maximise VERB ajst-31990 86 31 to to PART ajst-31990 86 32 improve improve VERB ajst-31990 86 33 6 6 NUM ajst-31990 86 34 diagnostic diagnostic ADJ ajst-31990 86 35 sensitivity sensitivity NOUN ajst-31990 86 36 ; ; PUNCT ajst-31990 86 37 false false ADJ ajst-31990 86 38 positive positive ADJ ajst-31990 86 39 ( ( PUNCT ajst-31990 86 40 fp fp NOUN ajst-31990 86 41 ) ) PUNCT ajst-31990 86 42 indicates indicate VERB ajst-31990 86 43 that that SCONJ ajst-31990 86 44 a a DET ajst-31990 86 45 healthy healthy ADJ ajst-31990 86 46 person person NOUN ajst-31990 86 47 is be AUX ajst-31990 86 48 misdiagnosed misdiagnose VERB ajst-31990 86 49 as as ADP ajst-31990 86 50 a a DET ajst-31990 86 51 patient patient NOUN ajst-31990 86 52 , , PUNCT ajst-31990 86 53 which which PRON ajst-31990 86 54 needs need VERB ajst-31990 86 55 to to PART ajst-31990 86 56 be be AUX ajst-31990 86 57 minimised minimise VERB ajst-31990 86 58 to to PART ajst-31990 86 59 avoid avoid VERB ajst-31990 86 60 over over ADP ajst-31990 86 61 - - PUNCT ajst-31990 86 62 treatment treatment NOUN ajst-31990 86 63 ; ; PUNCT ajst-31990 86 64 false false ADJ ajst-31990 86 65 negative negative ADJ ajst-31990 86 66 ( ( PUNCT ajst-31990 86 67 fn fn NOUN ajst-31990 86 68 ) ) PUNCT ajst-31990 86 69 indicates indicate VERB ajst-31990 86 70 that that SCONJ ajst-31990 86 71 a a DET ajst-31990 86 72 patient patient NOUN ajst-31990 86 73 is be AUX ajst-31990 86 74 under under ADV ajst-31990 86 75 - - PUNCT ajst-31990 86 76 diagnosed diagnose VERB ajst-31990 86 77 , , PUNCT ajst-31990 86 78 reflecting reflect VERB ajst-31990 86 79 the the DET ajst-31990 86 80 clinical clinical ADJ ajst-31990 86 81 risk risk NOUN ajst-31990 86 82 , , PUNCT ajst-31990 86 83 which which PRON ajst-31990 86 84 needs need VERB ajst-31990 86 85 to to PART ajst-31990 86 86 be be AUX ajst-31990 86 87 combined combine VERB ajst-31990 86 88 with with ADP ajst-31990 86 89 over over ADV ajst-31990 86 90 - - PUNCT ajst-31990 86 91 sampling sample VERB ajst-31990 86 92 techniques technique NOUN ajst-31990 86 93 or or CCONJ ajst-31990 86 94 model model NOUN ajst-31990 86 95 complexity complexity NOUN ajst-31990 86 96 enhancement enhancement NOUN ajst-31990 86 97 to to PART ajst-31990 86 98 reduce reduce VERB ajst-31990 86 99 the the DET ajst-31990 86 100 rate rate NOUN ajst-31990 86 101 of of ADP ajst-31990 86 102 under under NOUN ajst-31990 86 103 - - PUNCT ajst-31990 86 104 diagnosis diagnosis NOUN ajst-31990 86 105 ; ; PUNCT ajst-31990 86 106 and and CCONJ ajst-31990 86 107 true true ADJ ajst-31990 86 108 negative negative ADJ ajst-31990 86 109 ( ( PUNCT ajst-31990 86 110 tn tn NOUN ajst-31990 86 111 ) ) PUNCT ajst-31990 86 112 indicates indicate VERB ajst-31990 86 113 that that SCONJ ajst-31990 86 114 a a DET ajst-31990 86 115 sample sample NOUN ajst-31990 86 116 of of ADP ajst-31990 86 117 correctly correctly ADV ajst-31990 86 118 classified classified ADJ ajst-31990 86 119 healthy healthy ADJ ajst-31990 86 120 people people NOUN ajst-31990 86 121 , , PUNCT ajst-31990 86 122 reflecting reflect VERB ajst-31990 86 123 the the DET ajst-31990 86 124 model model NOUN ajst-31990 86 125 's 's PART ajst-31990 86 126 ability ability NOUN ajst-31990 86 127 to to PART ajst-31990 86 128 classify classify VERB ajst-31990 86 129 the the DET ajst-31990 86 130 negative negative ADJ ajst-31990 86 131 class class NOUN ajst-31990 86 132 as as ADP ajst-31990 86 133 a a DET ajst-31990 86 134 patient patient NOUN ajst-31990 86 135 . . PUNCT ajst-31990 87 1 tn tn PROPN ajst-31990 87 2 indicates indicate VERB ajst-31990 87 3 a a DET ajst-31990 87 4 correctly correctly ADV ajst-31990 87 5 classified classified ADJ ajst-31990 87 6 healthy healthy ADJ ajst-31990 87 7 sample sample NOUN ajst-31990 87 8 , , PUNCT ajst-31990 87 9 reflecting reflect VERB ajst-31990 87 10 the the DET ajst-31990 87 11 model model NOUN ajst-31990 87 12 's 's PART ajst-31990 87 13 ability ability NOUN ajst-31990 87 14 to to PART ajst-31990 87 15 distinguish distinguish VERB ajst-31990 87 16 between between ADP ajst-31990 87 17 negative negative ADJ ajst-31990 87 18 categories category NOUN ajst-31990 87 19 . . PUNCT ajst-31990 88 1 the the DET ajst-31990 88 2 matrix matrix NOUN ajst-31990 88 3 cross cross NOUN ajst-31990 88 4 - - NOUN ajst-31990 88 5 tabulates tabulate NOUN ajst-31990 88 6 predicted predict VERB ajst-31990 88 7 versus versus ADP ajst-31990 88 8 actual actual ADJ ajst-31990 88 9 correspondence correspondence NOUN ajst-31990 88 10 by by ADP ajst-31990 88 11 rows row NOUN ajst-31990 88 12 and and CCONJ ajst-31990 88 13 columns column NOUN ajst-31990 88 14 , , PUNCT ajst-31990 88 15 for for ADP ajst-31990 88 16 example example NOUN ajst-31990 88 17 , , PUNCT ajst-31990 88 18 in in ADP ajst-31990 88 19 medical medical ADJ ajst-31990 88 20 diagnosis diagnosis NOUN ajst-31990 88 21 , , PUNCT ajst-31990 88 22 where where SCONJ ajst-31990 88 23 the the DET ajst-31990 88 24 row row NOUN ajst-31990 88 25 data datum NOUN ajst-31990 88 26 represent represent VERB ajst-31990 88 27 the the DET ajst-31990 88 28 breakdown breakdown NOUN ajst-31990 88 29 of of ADP ajst-31990 88 30 those those PRON ajst-31990 88 31 who who PRON ajst-31990 88 32 are be AUX ajst-31990 88 33 actually actually ADV ajst-31990 88 34 ill ill ADJ ajst-31990 88 35 , , PUNCT ajst-31990 88 36 while while SCONJ ajst-31990 88 37 the the DET ajst-31990 88 38 column column NOUN ajst-31990 88 39 data datum NOUN ajst-31990 88 40 reflect reflect VERB ajst-31990 88 41 the the DET ajst-31990 88 42 actual actual ADJ ajst-31990 88 43 distribution distribution NOUN ajst-31990 88 44 in in ADP ajst-31990 88 45 the the DET ajst-31990 88 46 sample sample NOUN ajst-31990 88 47 predicted predict VERB ajst-31990 88 48 to to PART ajst-31990 88 49 be be AUX ajst-31990 88 50 ill ill ADJ ajst-31990 88 51 . . PUNCT ajst-31990 89 1 its its PRON ajst-31990 89 2 advantage advantage NOUN ajst-31990 89 3 is be AUX ajst-31990 89 4 that that SCONJ ajst-31990 89 5 it it PRON ajst-31990 89 6 provides provide VERB ajst-31990 89 7 a a DET ajst-31990 89 8 multi multi ADJ ajst-31990 89 9 - - ADJ ajst-31990 89 10 dimensional dimensional ADJ ajst-31990 89 11 perspective perspective NOUN ajst-31990 89 12 for for ADP ajst-31990 89 13 the the DET ajst-31990 89 14 model model NOUN ajst-31990 89 15 , , PUNCT ajst-31990 89 16 supports support VERB ajst-31990 89 17 the the DET ajst-31990 89 18 calculation calculation NOUN ajst-31990 89 19 of of ADP ajst-31990 89 20 precision precision NOUN ajst-31990 89 21 , , PUNCT ajst-31990 89 22 recall recall NOUN ajst-31990 89 23 and and CCONJ ajst-31990 89 24 other other ADJ ajst-31990 89 25 indicators indicator NOUN ajst-31990 89 26 , , PUNCT ajst-31990 89 27 and and CCONJ ajst-31990 89 28 is be AUX ajst-31990 89 29 especially especially ADV ajst-31990 89 30 suitable suitable ADJ ajst-31990 89 31 for for ADP ajst-31990 89 32 category category NOUN ajst-31990 89 33 imbalance imbalance NOUN ajst-31990 89 34 scenarios scenario NOUN ajst-31990 89 35 ( ( PUNCT ajst-31990 89 36 e.g. e.g. ADV ajst-31990 89 37 , , PUNCT ajst-31990 89 38 scarcity scarcity NOUN ajst-31990 89 39 of of ADP ajst-31990 89 40 patient patient ADJ ajst-31990 89 41 samples sample NOUN ajst-31990 89 42 in in ADP ajst-31990 89 43 medical medical ADJ ajst-31990 89 44 diagnosis diagnosis NOUN ajst-31990 89 45 ) ) PUNCT ajst-31990 89 46 . . PUNCT ajst-31990 90 1 2.6.3 2.6.3 NUM ajst-31990 90 2 classification classification NOUN ajst-31990 90 3 report report NOUN ajst-31990 90 4 . . PUNCT ajst-31990 91 1 classification classification NOUN ajst-31990 91 2 report report NOUN ajst-31990 91 3 as as ADP ajst-31990 91 4 a a DET ajst-31990 91 5 core core NOUN ajst-31990 91 6 tool tool NOUN ajst-31990 91 7 for for ADP ajst-31990 91 8 model model NOUN ajst-31990 91 9 performance performance NOUN ajst-31990 91 10 evaluation evaluation NOUN ajst-31990 91 11 , , PUNCT ajst-31990 91 12 which which PRON ajst-31990 91 13 quantifies quantify VERB ajst-31990 91 14 the the DET ajst-31990 91 15 classification classification NOUN ajst-31990 91 16 effect effect NOUN ajst-31990 91 17 by by ADP ajst-31990 91 18 structured structured ADJ ajst-31990 91 19 presentation presentation NOUN ajst-31990 91 20 of of ADP ajst-31990 91 21 multidimensional multidimensional ADJ ajst-31990 91 22 metrics metric NOUN ajst-31990 91 23 , , PUNCT ajst-31990 91 24 the the DET ajst-31990 91 25 core core NOUN ajst-31990 91 26 metrics metric NOUN ajst-31990 91 27 of of ADP ajst-31990 91 28 which which PRON ajst-31990 91 29 are be AUX ajst-31990 91 30 shown show VERB ajst-31990 91 31 in in ADP ajst-31990 91 32 table table NOUN ajst-31990 91 33 1 1 NUM ajst-31990 91 34 . . PUNCT ajst-31990 91 35 table table NOUN ajst-31990 91 36 1 1 NUM ajst-31990 91 37 . . NUM ajst-31990 91 38 formulas formula NOUN ajst-31990 91 39 and and CCONJ ajst-31990 91 40 meaning meaning NOUN ajst-31990 91 41 of of ADP ajst-31990 91 42 core core NOUN ajst-31990 91 43 indicators indicator NOUN ajst-31990 91 44 for for ADP ajst-31990 91 45 disaggregated disaggregate VERB ajst-31990 91 46 reporting reporting NOUN ajst-31990 91 47 norm norm NOUN ajst-31990 91 48 formula formula NOUN ajst-31990 91 49 medical medical ADJ ajst-31990 91 50 significance significance NOUN ajst-31990 91 51 precision precision NOUN ajst-31990 91 52 𝑇𝑃 𝑇𝑃 PROPN ajst-31990 91 53 𝑇𝑃 𝑇𝑃 PROPN ajst-31990 91 54 + + CCONJ ajst-31990 91 55 𝐹𝑃 𝐹𝑃 PROPN ajst-31990 91 56 reduced reduce VERB ajst-31990 91 57 misdiagnosis misdiagnosis NOUN ajst-31990 91 58 ( ( PUNCT ajst-31990 91 59 fewer few ADJ ajst-31990 91 60 healthy healthy ADJ ajst-31990 91 61 people people NOUN ajst-31990 91 62 misdiagnosed misdiagnose VERB ajst-31990 91 63 ) ) PUNCT ajst-31990 91 64 recall recall NOUN ajst-31990 91 65 rate rate NOUN ajst-31990 91 66 𝑇𝑃 𝑇𝑃 PROPN ajst-31990 91 67 𝑇𝑃 𝑇𝑃 PROPN ajst-31990 91 68 + + CCONJ ajst-31990 91 69 𝐹𝑁 𝐹𝑁 PROPN ajst-31990 91 70 reducing reduce VERB ajst-31990 91 71 underdiagnosis underdiagnosis NOUN ajst-31990 91 72 ( ( PUNCT ajst-31990 91 73 ensuring ensure VERB ajst-31990 91 74 patients patient NOUN ajst-31990 91 75 are be AUX ajst-31990 91 76 detected detect VERB ajst-31990 91 77 ) ) PUNCT ajst-31990 91 78 f1 f1 NOUN ajst-31990 91 79 score score NOUN ajst-31990 91 80 2Γ— 2Γ— NUM ajst-31990 91 81 π‘ƒπ‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›Γ—π‘…π‘’π‘π‘Žπ‘™π‘™ π‘ƒπ‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›Γ—π‘…π‘’π‘π‘Žπ‘™π‘™ PROPN ajst-31990 91 82 π‘ƒπ‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›+π‘…π‘’π‘π‘Žπ‘™π‘™ π‘ƒπ‘Ÿπ‘’π‘π‘–π‘ π‘–π‘œπ‘›+π‘…π‘’π‘π‘Žπ‘™π‘™ ADJ ajst-31990 91 83 comprehensive comprehensive ADJ ajst-31990 91 84 indicators indicator NOUN ajst-31990 91 85 to to PART ajst-31990 91 86 balance balance VERB ajst-31990 91 87 misdiagnosis misdiagnosis NOUN ajst-31990 91 88 and and CCONJ ajst-31990 91 89 underdiagnosis underdiagnosis VERB ajst-31990 91 90 additional additional ADJ ajst-31990 91 91 metrics metric NOUN ajst-31990 91 92 such such ADJ ajst-31990 91 93 as as ADP ajst-31990 91 94 support support NOUN ajst-31990 91 95 provide provide VERB ajst-31990 91 96 the the DET ajst-31990 91 97 true true ADJ ajst-31990 91 98 sample sample NOUN ajst-31990 91 99 size size NOUN ajst-31990 91 100 for for ADP ajst-31990 91 101 each each DET ajst-31990 91 102 category category NOUN ajst-31990 91 103 , , PUNCT ajst-31990 91 104 while while SCONJ ajst-31990 91 105 macro macro ADJ ajst-31990 91 106 avg avg NOUN ajst-31990 91 107 and and CCONJ ajst-31990 91 108 weighted weight VERB ajst-31990 91 109 avg avg NOUN ajst-31990 91 110 assess assess VERB ajst-31990 91 111 the the DET ajst-31990 91 112 model model NOUN ajst-31990 91 113 generalisation generalisation NOUN ajst-31990 91 114 ability ability NOUN ajst-31990 91 115 from from ADP ajst-31990 91 116 an an DET ajst-31990 91 117 equality equality NOUN ajst-31990 91 118 perspective perspective NOUN ajst-31990 91 119 and and CCONJ ajst-31990 91 120 data datum NOUN ajst-31990 91 121 distribution distribution NOUN ajst-31990 91 122 weighting weight VERB ajst-31990 91 123 perspective perspective NOUN ajst-31990 91 124 respectively.for respectively.for ADP ajst-31990 91 125 example example NOUN ajst-31990 91 126 , , PUNCT ajst-31990 91 127 in in ADP ajst-31990 91 128 medical medical ADJ ajst-31990 91 129 diagnosis diagnosis NOUN ajst-31990 91 130 , , PUNCT ajst-31990 91 131 the the DET ajst-31990 91 132 weighted weight VERB ajst-31990 91 133 average average NOUN ajst-31990 91 134 focuses focus VERB ajst-31990 91 135 more more ADJ ajst-31990 91 136 on on ADP ajst-31990 91 137 the the DET ajst-31990 91 138 contribution contribution NOUN ajst-31990 91 139 of of ADP ajst-31990 91 140 the the DET ajst-31990 91 141 majority majority NOUN ajst-31990 91 142 of of ADP ajst-31990 91 143 the the DET ajst-31990 91 144 class class NOUN ajst-31990 91 145 samples sample NOUN ajst-31990 91 146 , , PUNCT ajst-31990 91 147 while while SCONJ ajst-31990 91 148 the the DET ajst-31990 91 149 macro macro ADJ ajst-31990 91 150 average average NOUN ajst-31990 91 151 is be AUX ajst-31990 91 152 suitable suitable ADJ ajst-31990 91 153 for for ADP ajst-31990 91 154 scenarios scenario NOUN ajst-31990 91 155 that that PRON ajst-31990 91 156 require require VERB ajst-31990 91 157 a a DET ajst-31990 91 158 balanced balanced ADJ ajst-31990 91 159 focus focus NOUN ajst-31990 91 160 on on ADP ajst-31990 91 161 the the DET ajst-31990 91 162 subclasses subclass NOUN ajst-31990 91 163 . . PUNCT ajst-31990 92 1 3 3 X ajst-31990 92 2 . . X ajst-31990 92 3 analysis analysis NOUN ajst-31990 92 4 of of ADP ajst-31990 92 5 the the DET ajst-31990 92 6 results result NOUN ajst-31990 92 7 of of ADP ajst-31990 92 8 the the DET ajst-31990 92 9 study study NOUN ajst-31990 92 10 2.7 2.7 NUM ajst-31990 92 11 . . PUNCT ajst-31990 93 1 roc roc NOUN ajst-31990 93 2 - - PUNCT ajst-31990 93 3 auc auc VERB ajst-31990 93 4 the the DET ajst-31990 93 5 data datum NOUN ajst-31990 93 6 from from ADP ajst-31990 93 7 the the DET ajst-31990 93 8 roc roc NOUN ajst-31990 93 9 - - PUNCT ajst-31990 93 10 auc auc NOUN ajst-31990 93 11 outputs output NOUN ajst-31990 93 12 of of ADP ajst-31990 93 13 the the DET ajst-31990 93 14 three three NUM ajst-31990 93 15 models model NOUN ajst-31990 93 16 are be AUX ajst-31990 93 17 shown show VERB ajst-31990 93 18 in in ADP ajst-31990 93 19 table table NOUN ajst-31990 93 20 2 2 NUM ajst-31990 93 21 . . PUNCT ajst-31990 93 22 table table NOUN ajst-31990 93 23 2 2 NUM ajst-31990 93 24 . . PUNCT ajst-31990 93 25 comparison comparison NOUN ajst-31990 93 26 of of ADP ajst-31990 93 27 the the DET ajst-31990 93 28 performance performance NOUN ajst-31990 93 29 of of ADP ajst-31990 93 30 the the DET ajst-31990 93 31 three three NUM ajst-31990 93 32 models model NOUN ajst-31990 93 33 with with ADP ajst-31990 93 34 respect respect NOUN ajst-31990 93 35 to to PART ajst-31990 93 36 auc auc VERB ajst-31990 93 37 training training NOUN ajst-31990 93 38 accuracy accuracy NOUN ajst-31990 93 39 verification verification NOUN ajst-31990 93 40 accuracy accuracy NOUN ajst-31990 93 41 logistic logistic ADJ ajst-31990 93 42 regression regression NOUN ajst-31990 93 43 82.92 82.92 NUM ajst-31990 93 44 % % NOUN ajst-31990 93 45 81.47 81.47 NUM ajst-31990 93 46 % % NOUN ajst-31990 93 47 xgboost xgboost NOUN ajst-31990 94 1 0.99β‰ˆ100 0.99β‰ˆ100 PROPN ajst-31990 94 2 % % PROPN ajst-31990 94 3 78.57 78.57 NUM ajst-31990 94 4 % % NOUN ajst-31990 94 5 svm svm NOUN ajst-31990 94 6 71.05 71.05 NUM ajst-31990 94 7 % % NOUN ajst-31990 94 8 75.87 75.87 NUM ajst-31990 94 9 % % NOUN ajst-31990 94 10 the the DET ajst-31990 94 11 assessment assessment NOUN ajst-31990 94 12 of of ADP ajst-31990 94 13 the the DET ajst-31990 94 14 model model NOUN ajst-31990 94 15 performance performance NOUN ajst-31990 94 16 shows show VERB ajst-31990 94 17 that that SCONJ ajst-31990 94 18 the the DET ajst-31990 94 19 logistic logistic ADJ ajst-31990 94 20 regression regression NOUN ajst-31990 94 21 model model NOUN ajst-31990 94 22 has have VERB ajst-31990 94 23 an an DET ajst-31990 94 24 accuracy accuracy NOUN ajst-31990 94 25 gap gap NOUN ajst-31990 94 26 of of ADP ajst-31990 94 27 only only ADV ajst-31990 94 28 1.45 1.45 NUM ajst-31990 94 29 % % NOUN ajst-31990 94 30 between between ADP ajst-31990 94 31 the the DET ajst-31990 94 32 training training NOUN ajst-31990 94 33 set set NOUN ajst-31990 94 34 ( ( PUNCT ajst-31990 94 35 82.92 82.92 NUM ajst-31990 94 36 % % NOUN ajst-31990 94 37 ) ) PUNCT ajst-31990 94 38 and and CCONJ ajst-31990 94 39 the the DET ajst-31990 94 40 validation validation NOUN ajst-31990 94 41 set set NOUN ajst-31990 94 42 ( ( PUNCT ajst-31990 94 43 81.47 81.47 NUM ajst-31990 94 44 % % NOUN ajst-31990 94 45 ) ) PUNCT ajst-31990 94 46 , , PUNCT ajst-31990 94 47 indicating indicate VERB ajst-31990 94 48 that that SCONJ ajst-31990 94 49 it it PRON ajst-31990 94 50 does do AUX ajst-31990 94 51 not not PART ajst-31990 94 52 show show VERB ajst-31990 94 53 significant significant ADJ ajst-31990 94 54 overfitting overfitting NOUN ajst-31990 94 55 . . PUNCT ajst-31990 95 1 the the DET ajst-31990 95 2 model model NOUN ajst-31990 95 3 shows show VERB ajst-31990 95 4 better well ADJ ajst-31990 95 5 generalisation generalisation NOUN ajst-31990 95 6 ability ability NOUN ajst-31990 95 7 due due ADP ajst-31990 95 8 to to ADP ajst-31990 95 9 the the DET ajst-31990 95 10 low low ADJ ajst-31990 95 11 complexity complexity NOUN ajst-31990 95 12 characteristics characteristic NOUN ajst-31990 95 13 of of ADP ajst-31990 95 14 logistic logistic ADJ ajst-31990 95 15 regression regression NOUN ajst-31990 95 16 itself itself PRON ajst-31990 95 17 and and CCONJ ajst-31990 95 18 moderate moderate ADJ ajst-31990 95 19 regularisation regularisation NOUN ajst-31990 95 20 strategy strategy NOUN ajst-31990 95 21 , , PUNCT ajst-31990 95 22 but but CCONJ ajst-31990 95 23 there there PRON ajst-31990 95 24 is be VERB ajst-31990 95 25 still still ADV ajst-31990 95 26 room room NOUN ajst-31990 95 27 for for ADP ajst-31990 95 28 improvement improvement NOUN ajst-31990 95 29 in in ADP ajst-31990 95 30 the the DET ajst-31990 95 31 overall overall ADJ ajst-31990 95 32 performance performance NOUN ajst-31990 95 33 , , PUNCT ajst-31990 95 34 and and CCONJ ajst-31990 95 35 the the DET ajst-31990 95 36 model model NOUN ajst-31990 95 37 potential potential NOUN ajst-31990 95 38 may may AUX ajst-31990 95 39 need need VERB ajst-31990 95 40 to to PART ajst-31990 95 41 be be AUX ajst-31990 95 42 further far ADV ajst-31990 95 43 explored explore VERB ajst-31990 95 44 through through ADP ajst-31990 95 45 feature feature NOUN ajst-31990 95 46 engineering engineering NOUN ajst-31990 95 47 optimisation optimisation NOUN ajst-31990 95 48 or or CCONJ ajst-31990 95 49 parameter parameter NOUN ajst-31990 95 50 tuning tuning NOUN ajst-31990 95 51 . . PUNCT ajst-31990 96 1 7 7 NUM ajst-31990 97 1 the the DET ajst-31990 97 2 xgboost xgboost PROPN ajst-31990 97 3 model model NOUN ajst-31990 97 4 performs perform VERB ajst-31990 97 5 nearly nearly ADV ajst-31990 97 6 perfectly perfectly ADV ajst-31990 97 7 on on ADP ajst-31990 97 8 the the DET ajst-31990 97 9 training training NOUN ajst-31990 97 10 set set NOUN ajst-31990 97 11 ( ( PUNCT ajst-31990 97 12 99 99 NUM ajst-31990 97 13 % % NOUN ajst-31990 97 14 accuracy accuracy NOUN ajst-31990 97 15 ) ) PUNCT ajst-31990 97 16 , , PUNCT ajst-31990 97 17 but but CCONJ ajst-31990 97 18 the the DET ajst-31990 97 19 accuracy accuracy NOUN ajst-31990 97 20 on on ADP ajst-31990 97 21 the the DET ajst-31990 97 22 validation validation NOUN ajst-31990 97 23 set set VERB ajst-31990 97 24 plummets plummet NOUN ajst-31990 97 25 to to ADP ajst-31990 97 26 78.57 78.57 NUM ajst-31990 97 27 % % NOUN ajst-31990 97 28 , , PUNCT ajst-31990 97 29 a a DET ajst-31990 97 30 gap gap NOUN ajst-31990 97 31 of of ADP ajst-31990 97 32 21.43 21.43 NUM ajst-31990 97 33 % % NOUN ajst-31990 97 34 , , PUNCT ajst-31990 97 35 showing show VERB ajst-31990 97 36 typical typical ADJ ajst-31990 97 37 characteristics characteristic NOUN ajst-31990 97 38 of of ADP ajst-31990 97 39 severe severe ADJ ajst-31990 97 40 overfitting overfitting NOUN ajst-31990 97 41 . . PUNCT ajst-31990 98 1 this this PRON ajst-31990 98 2 may may AUX ajst-31990 98 3 stem stem VERB ajst-31990 98 4 from from ADP ajst-31990 98 5 excessive excessive ADJ ajst-31990 98 6 model model NOUN ajst-31990 98 7 complexity complexity NOUN ajst-31990 98 8 and and CCONJ ajst-31990 98 9 lack lack NOUN ajst-31990 98 10 of of ADP ajst-31990 98 11 regularisation regularisation NOUN ajst-31990 98 12 , , PUNCT ajst-31990 98 13 leading lead VERB ajst-31990 98 14 to to ADP ajst-31990 98 15 overfitting overfitting NOUN ajst-31990 98 16 of of ADP ajst-31990 98 17 the the DET ajst-31990 98 18 model model NOUN ajst-31990 98 19 to to ADP ajst-31990 98 20 the the DET ajst-31990 98 21 training training NOUN ajst-31990 98 22 noise noise NOUN ajst-31990 98 23 . . PUNCT ajst-31990 99 1 this this PRON ajst-31990 99 2 is be AUX ajst-31990 99 3 in in ADP ajst-31990 99 4 line line NOUN ajst-31990 99 5 with with ADP ajst-31990 99 6 xgboost xgboost PROPN ajst-31990 99 7 's 's PART ajst-31990 99 8 vulnerability vulnerability NOUN ajst-31990 99 9 to to PART ajst-31990 99 10 feature feature VERB ajst-31990 99 11 covariance covariance NOUN ajst-31990 99 12 as as ADP ajst-31990 99 13 a a DET ajst-31990 99 14 complex complex ADJ ajst-31990 99 15 integrated integrated ADJ ajst-31990 99 16 model model NOUN ajst-31990 99 17 , , PUNCT ajst-31990 99 18 and and CCONJ ajst-31990 99 19 overfitting overfitte VERB ajst-31990 99 20 needs need NOUN ajst-31990 99 21 to to PART ajst-31990 99 22 be be AUX ajst-31990 99 23 mitigated mitigate VERB ajst-31990 99 24 by by ADP ajst-31990 99 25 regularisation regularisation NOUN ajst-31990 99 26 terms term NOUN ajst-31990 99 27 or or CCONJ ajst-31990 99 28 feature feature NOUN ajst-31990 99 29 selection selection NOUN ajst-31990 99 30 techniques technique NOUN ajst-31990 99 31 . . PUNCT ajst-31990 100 1 svm svm PROPN ajst-31990 100 2 , , PUNCT ajst-31990 100 3 on on ADP ajst-31990 100 4 the the DET ajst-31990 100 5 other other ADJ ajst-31990 100 6 hand hand NOUN ajst-31990 100 7 , , PUNCT ajst-31990 100 8 presents present VERB ajst-31990 100 9 the the DET ajst-31990 100 10 peculiar peculiar ADJ ajst-31990 100 11 phenomenon phenomenon NOUN ajst-31990 100 12 that that SCONJ ajst-31990 100 13 the the DET ajst-31990 100 14 validation validation NOUN ajst-31990 100 15 accuracy accuracy NOUN ajst-31990 100 16 ( ( PUNCT ajst-31990 100 17 75.87 75.87 NUM ajst-31990 100 18 % % NOUN ajst-31990 100 19 ) ) PUNCT ajst-31990 100 20 is be AUX ajst-31990 100 21 higher high ADJ ajst-31990 100 22 than than ADP ajst-31990 100 23 the the DET ajst-31990 100 24 training training NOUN ajst-31990 100 25 accuracy accuracy NOUN ajst-31990 100 26 ( ( PUNCT ajst-31990 100 27 71.05 71.05 NUM ajst-31990 100 28 % % NOUN ajst-31990 100 29 ) ) PUNCT ajst-31990 100 30 , , PUNCT ajst-31990 100 31 which which PRON ajst-31990 100 32 may may AUX ajst-31990 100 33 stem stem VERB ajst-31990 100 34 from from ADP ajst-31990 100 35 underestimation underestimation NOUN ajst-31990 100 36 of of ADP ajst-31990 100 37 the the DET ajst-31990 100 38 validation validation NOUN ajst-31990 100 39 error error NOUN ajst-31990 100 40 due due ADP ajst-31990 100 41 to to ADP ajst-31990 100 42 the the DET ajst-31990 100 43 randomness randomness NOUN ajst-31990 100 44 of of ADP ajst-31990 100 45 the the DET ajst-31990 100 46 data data NOUN ajst-31990 100 47 partitioning partition VERB ajst-31990 100 48 or or CCONJ ajst-31990 100 49 the the DET ajst-31990 100 50 higher high ADJ ajst-31990 100 51 strength strength NOUN ajst-31990 100 52 of of ADP ajst-31990 100 53 the the DET ajst-31990 100 54 model model NOUN ajst-31990 100 55 regularisation regularisation NOUN ajst-31990 100 56 . . PUNCT ajst-31990 101 1 the the DET ajst-31990 101 2 overall overall ADJ ajst-31990 101 3 low low ADJ ajst-31990 101 4 performance performance NOUN ajst-31990 101 5 may may AUX ajst-31990 101 6 be be AUX ajst-31990 101 7 related relate VERB ajst-31990 101 8 to to ADP ajst-31990 101 9 the the DET ajst-31990 101 10 improper improper ADJ ajst-31990 101 11 selection selection NOUN ajst-31990 101 12 of of ADP ajst-31990 101 13 kernel kernel NOUN ajst-31990 101 14 functions function NOUN ajst-31990 101 15 or or CCONJ ajst-31990 101 16 insufficient insufficient ADJ ajst-31990 101 17 parameter parameter NOUN ajst-31990 101 18 tuning tuning NOUN ajst-31990 101 19 , , PUNCT ajst-31990 101 20 and and CCONJ ajst-31990 101 21 it it PRON ajst-31990 101 22 is be AUX ajst-31990 101 23 necessary necessary ADJ ajst-31990 101 24 to to PART ajst-31990 101 25 optimise optimise VERB ajst-31990 101 26 the the DET ajst-31990 101 27 kernel kernel PROPN ajst-31990 101 28 functions function NOUN ajst-31990 101 29 and and CCONJ ajst-31990 101 30 hyperparameter hyperparameter NOUN ajst-31990 101 31 combinations combination NOUN ajst-31990 101 32 by by ADP ajst-31990 101 33 combining combine VERB ajst-31990 101 34 cross cross NOUN ajst-31990 101 35 - - ADJ ajst-31990 101 36 validation validation ADJ ajst-31990 101 37 and and CCONJ ajst-31990 101 38 grid grid NOUN ajst-31990 101 39 search search NOUN ajst-31990 101 40 . . PUNCT ajst-31990 102 1 2.8 2.8 NUM ajst-31990 102 2 . . PUNCT ajst-31990 102 3 confusion confusion NOUN ajst-31990 102 4 matrix matrix NOUN ajst-31990 102 5 and and CCONJ ajst-31990 102 6 classification classification NOUN ajst-31990 102 7 report report NOUN ajst-31990 102 8 the the DET ajst-31990 102 9 results result NOUN ajst-31990 102 10 of of ADP ajst-31990 102 11 the the DET ajst-31990 102 12 confusion confusion NOUN ajst-31990 102 13 matrix matrix NOUN ajst-31990 102 14 analysis analysis NOUN ajst-31990 102 15 of of ADP ajst-31990 102 16 the the DET ajst-31990 102 17 logistic logistic ADJ ajst-31990 102 18 regression regression NOUN ajst-31990 102 19 model model NOUN ajst-31990 102 20 on on ADP ajst-31990 102 21 the the DET ajst-31990 102 22 validation validation NOUN ajst-31990 102 23 set set NOUN ajst-31990 102 24 are be AUX ajst-31990 102 25 shown show VERB ajst-31990 102 26 in in ADP ajst-31990 102 27 figure figure NOUN ajst-31990 102 28 2 2 NUM ajst-31990 102 29 , , PUNCT ajst-31990 102 30 which which PRON ajst-31990 102 31 shows show VERB ajst-31990 102 32 that that SCONJ ajst-31990 102 33 there there PRON ajst-31990 102 34 is be VERB ajst-31990 102 35 a a DET ajst-31990 102 36 significant significant ADJ ajst-31990 102 37 difference difference NOUN ajst-31990 102 38 in in ADP ajst-31990 102 39 the the DET ajst-31990 102 40 model model NOUN ajst-31990 102 41 's 's PART ajst-31990 102 42 ability ability NOUN ajst-31990 102 43 to to PART ajst-31990 102 44 identify identify VERB ajst-31990 102 45 positive positive ADJ ajst-31990 102 46 ( ( PUNCT ajst-31990 102 47 1.0 1.0 NUM ajst-31990 102 48 ) ) PUNCT ajst-31990 102 49 and and CCONJ ajst-31990 102 50 negative negative ADJ ajst-31990 102 51 ( ( PUNCT ajst-31990 102 52 0.0 0.0 NUM ajst-31990 102 53 ) ) PUNCT ajst-31990 102 54 classes class NOUN ajst-31990 102 55 . . PUNCT ajst-31990 103 1 for for ADP ajst-31990 103 2 the the DET ajst-31990 103 3 positive positive ADJ ajst-31990 103 4 category category NOUN ajst-31990 103 5 , , PUNCT ajst-31990 103 6 the the DET ajst-31990 103 7 model model NOUN ajst-31990 103 8 correctly correctly ADV ajst-31990 103 9 identified identify VERB ajst-31990 103 10 35 35 NUM ajst-31990 103 11 positive positive ADJ ajst-31990 103 12 samples sample NOUN ajst-31990 103 13 ( ( PUNCT ajst-31990 103 14 tp tp NOUN ajst-31990 103 15 ) ) PUNCT ajst-31990 103 16 , , PUNCT ajst-31990 103 17 but but CCONJ ajst-31990 103 18 also also ADV ajst-31990 103 19 misclassified misclassifie VERB ajst-31990 103 20 4 4 NUM ajst-31990 103 21 positive positive ADJ ajst-31990 103 22 samples sample NOUN ajst-31990 103 23 as as ADP ajst-31990 103 24 negative negative ADJ ajst-31990 103 25 ( ( PUNCT ajst-31990 103 26 fn fn NOUN ajst-31990 103 27 ) ) PUNCT ajst-31990 103 28 ; ; PUNCT ajst-31990 103 29 for for ADP ajst-31990 103 30 the the DET ajst-31990 103 31 negative negative ADJ ajst-31990 103 32 category category NOUN ajst-31990 103 33 , , PUNCT ajst-31990 103 34 the the DET ajst-31990 103 35 model model NOUN ajst-31990 103 36 correctly correctly ADV ajst-31990 103 37 classified classify VERB ajst-31990 103 38 10 10 NUM ajst-31990 103 39 negative negative ADJ ajst-31990 103 40 samples sample NOUN ajst-31990 103 41 ( ( PUNCT ajst-31990 103 42 tn tn NOUN ajst-31990 103 43 ) ) PUNCT ajst-31990 103 44 , , PUNCT ajst-31990 103 45 but but CCONJ ajst-31990 103 46 2 2 NUM ajst-31990 103 47 negative negative ADJ ajst-31990 103 48 samples sample NOUN ajst-31990 103 49 were be AUX ajst-31990 103 50 incorrectly incorrectly ADV ajst-31990 103 51 predicted predict VERB ajst-31990 103 52 as as ADP ajst-31990 103 53 positive positive ADJ ajst-31990 103 54 ( ( PUNCT ajst-31990 103 55 fp fp NOUN ajst-31990 103 56 ) ) PUNCT ajst-31990 103 57 . . PUNCT ajst-31990 104 1 this this DET ajst-31990 104 2 result result NOUN ajst-31990 104 3 indicates indicate VERB ajst-31990 104 4 that that SCONJ ajst-31990 104 5 the the DET ajst-31990 104 6 model model NOUN ajst-31990 104 7 identifies identify VERB ajst-31990 104 8 the the DET ajst-31990 104 9 positive positive ADJ ajst-31990 104 10 classes class NOUN ajst-31990 104 11 better well ADV ajst-31990 104 12 ( ( PUNCT ajst-31990 104 13 recall recall VERB ajst-31990 104 14 = = SYM ajst-31990 104 15 35/(35 35/(35 NUM ajst-31990 104 16 + + SYM ajst-31990 104 17 4 4 NUM ajst-31990 104 18 ) ) PUNCT ajst-31990 104 19 = = PUNCT ajst-31990 104 20 89.7 89.7 NUM ajst-31990 104 21 % % NOUN ajst-31990 104 22 ) ) PUNCT ajst-31990 104 23 , , PUNCT ajst-31990 104 24 but but CCONJ ajst-31990 104 25 there there PRON ajst-31990 104 26 is be VERB ajst-31990 104 27 some some DET ajst-31990 104 28 degree degree NOUN ajst-31990 104 29 of of ADP ajst-31990 104 30 misclassification misclassification NOUN ajst-31990 104 31 for for ADP ajst-31990 104 32 the the DET ajst-31990 104 33 negative negative ADJ ajst-31990 104 34 classes class NOUN ajst-31990 104 35 ( ( PUNCT ajst-31990 104 36 precision precision NOUN ajst-31990 104 37 = = SYM ajst-31990 104 38 10/(10 10/(10 NUM ajst-31990 104 39 + + NOUN ajst-31990 104 40 2 2 NUM ajst-31990 104 41 ) ) PUNCT ajst-31990 104 42 = = SYM ajst-31990 104 43 83.3 83.3 NUM ajst-31990 104 44 % % NOUN ajst-31990 104 45 ) ) PUNCT ajst-31990 104 46 . . PUNCT ajst-31990 105 1 this this DET ajst-31990 105 2 logistic logistic ADJ ajst-31990 105 3 regression regression NOUN ajst-31990 105 4 type type NOUN ajst-31990 105 5 demonstrates demonstrate VERB ajst-31990 105 6 a a DET ajst-31990 105 7 stronger strong ADJ ajst-31990 105 8 ability ability NOUN ajst-31990 105 9 to to PART ajst-31990 105 10 capture capture VERB ajst-31990 105 11 positive positive ADJ ajst-31990 105 12 class class NOUN ajst-31990 105 13 samples sample NOUN ajst-31990 105 14 , , PUNCT ajst-31990 105 15 but but CCONJ ajst-31990 105 16 there there PRON ajst-31990 105 17 is be VERB ajst-31990 105 18 still still ADV ajst-31990 105 19 room room NOUN ajst-31990 105 20 for for ADP ajst-31990 105 21 improvement improvement NOUN ajst-31990 105 22 in in ADP ajst-31990 105 23 discriminative discriminative NOUN ajst-31990 105 24 performance performance NOUN ajst-31990 105 25 for for ADP ajst-31990 105 26 negative negative ADJ ajst-31990 105 27 class class NOUN ajst-31990 105 28 samples sample NOUN ajst-31990 105 29 . . PUNCT ajst-31990 106 1 figure figure NOUN ajst-31990 106 2 2 2 NUM ajst-31990 106 3 . . PUNCT ajst-31990 106 4 confusion confusion NOUN ajst-31990 106 5 matrix matrix NOUN ajst-31990 106 6 for for ADP ajst-31990 106 7 logistic logistic ADJ ajst-31990 106 8 regression regression NOUN ajst-31990 106 9 ( ( PUNCT ajst-31990 106 10 picture picture NOUN ajst-31990 106 11 credit credit NOUN ajst-31990 106 12 : : PUNCT ajst-31990 106 13 original original ADJ ajst-31990 106 14 ) ) PUNCT ajst-31990 106 15 the the DET ajst-31990 106 16 classification classification NOUN ajst-31990 106 17 report report NOUN ajst-31990 106 18 for for ADP ajst-31990 106 19 logistic logistic ADJ ajst-31990 106 20 regression regression NOUN ajst-31990 106 21 is be AUX ajst-31990 106 22 shown show VERB ajst-31990 106 23 in in ADP ajst-31990 106 24 table table NOUN ajst-31990 106 25 3 3 NUM ajst-31990 106 26 . . PUNCT ajst-31990 107 1 the the DET ajst-31990 107 2 overall overall ADJ ajst-31990 107 3 accuracy accuracy NOUN ajst-31990 107 4 of of ADP ajst-31990 107 5 the the DET ajst-31990 107 6 svm svm ADJ ajst-31990 107 7 model model NOUN ajst-31990 107 8 was be AUX ajst-31990 107 9 88 88 NUM ajst-31990 107 10 % % NOUN ajst-31990 107 11 , , PUNCT ajst-31990 107 12 with with ADP ajst-31990 107 13 the the DET ajst-31990 107 14 positive positive ADJ ajst-31990 107 15 category category NOUN ajst-31990 107 16 ( ( PUNCT ajst-31990 107 17 1.0 1.0 NUM ajst-31990 107 18 ) ) PUNCT ajst-31990 107 19 being be AUX ajst-31990 107 20 significantly significantly ADV ajst-31990 107 21 better well ADJ ajst-31990 107 22 than than ADP ajst-31990 107 23 the the DET ajst-31990 107 24 negative negative ADJ ajst-31990 107 25 category category NOUN ajst-31990 107 26 ( ( PUNCT ajst-31990 107 27 0.0 0.0 NUM ajst-31990 107 28 ) ) PUNCT ajst-31990 107 29 in in ADP ajst-31990 107 30 terms term NOUN ajst-31990 107 31 of of ADP ajst-31990 107 32 identification identification NOUN ajst-31990 107 33 . . PUNCT ajst-31990 108 1 for for ADP ajst-31990 108 2 the the DET ajst-31990 108 3 positive positive ADJ ajst-31990 108 4 class class NOUN ajst-31990 108 5 , , PUNCT ajst-31990 108 6 the the DET ajst-31990 108 7 precision precision NOUN ajst-31990 108 8 ( ( PUNCT ajst-31990 108 9 90 90 NUM ajst-31990 108 10 % % NOUN ajst-31990 108 11 ) ) PUNCT ajst-31990 108 12 , , PUNCT ajst-31990 108 13 recall recall INTJ ajst-31990 108 14 ( ( PUNCT ajst-31990 108 15 95 95 NUM ajst-31990 108 16 % % NOUN ajst-31990 108 17 ) ) PUNCT ajst-31990 108 18 and and CCONJ ajst-31990 108 19 f1 f1 PROPN ajst-31990 108 20 score score NOUN ajst-31990 108 21 ( ( PUNCT ajst-31990 108 22 0.92 0.92 NUM ajst-31990 108 23 ) ) PUNCT ajst-31990 108 24 are be AUX ajst-31990 108 25 at at ADP ajst-31990 108 26 a a DET ajst-31990 108 27 high high ADJ ajst-31990 108 28 level level NOUN ajst-31990 108 29 , , PUNCT ajst-31990 108 30 indicating indicate VERB ajst-31990 108 31 that that SCONJ ajst-31990 108 32 the the DET ajst-31990 108 33 model model NOUN ajst-31990 108 34 is be AUX ajst-31990 108 35 effective effective ADJ ajst-31990 108 36 in in ADP ajst-31990 108 37 capturing capture VERB ajst-31990 108 38 positive positive ADJ ajst-31990 108 39 class class NOUN ajst-31990 108 40 samples sample NOUN ajst-31990 108 41 ; ; PUNCT ajst-31990 108 42 whereas whereas SCONJ ajst-31990 108 43 the the DET ajst-31990 108 44 negative negative ADJ ajst-31990 108 45 category category NOUN ajst-31990 108 46 ( ( PUNCT ajst-31990 108 47 0.0 0.0 NUM ajst-31990 108 48 ) ) PUNCT ajst-31990 108 49 has have VERB ajst-31990 108 50 a a DET ajst-31990 108 51 lower low ADJ ajst-31990 108 52 recall recall NOUN ajst-31990 108 53 ( ( PUNCT ajst-31990 108 54 71 71 NUM ajst-31990 108 55 % % NOUN ajst-31990 108 56 ) ) PUNCT ajst-31990 108 57 and and CCONJ ajst-31990 108 58 some some DET ajst-31990 108 59 risk risk NOUN ajst-31990 108 60 of of ADP ajst-31990 108 61 underestimation underestimation NOUN ajst-31990 108 62 , , PUNCT ajst-31990 108 63 its its PRON ajst-31990 108 64 precision precision NOUN ajst-31990 108 65 ( ( PUNCT ajst-31990 108 66 83 83 NUM ajst-31990 108 67 % % NOUN ajst-31990 108 68 ) ) PUNCT ajst-31990 108 69 still still ADV ajst-31990 108 70 indicates indicate VERB ajst-31990 108 71 a a DET ajst-31990 108 72 manageable manageable ADJ ajst-31990 108 73 proportion proportion NOUN ajst-31990 108 74 of of ADP ajst-31990 108 75 misclassifications misclassification NOUN ajst-31990 108 76 predicted predict VERB ajst-31990 108 77 to to PART ajst-31990 108 78 be be AUX ajst-31990 108 79 negative negative ADJ ajst-31990 108 80 . . PUNCT ajst-31990 109 1 further further ADJ ajst-31990 109 2 analysis analysis NOUN ajst-31990 109 3 revealed reveal VERB ajst-31990 109 4 that that SCONJ ajst-31990 109 5 due due ADP ajst-31990 109 6 to to ADP ajst-31990 109 7 the the DET ajst-31990 109 8 category category NOUN ajst-31990 109 9 imbalance imbalance NOUN ajst-31990 109 10 in in ADP ajst-31990 109 11 the the DET ajst-31990 109 12 data datum NOUN ajst-31990 109 13 ( ( PUNCT ajst-31990 109 14 14 14 NUM ajst-31990 109 15 cases case NOUN ajst-31990 109 16 for for ADP ajst-31990 109 17 negative negative ADJ ajst-31990 109 18 categories category NOUN ajst-31990 109 19 and and CCONJ ajst-31990 109 20 37 37 NUM ajst-31990 109 21 cases case NOUN ajst-31990 109 22 for for ADP ajst-31990 109 23 positive positive ADJ ajst-31990 109 24 categories category NOUN ajst-31990 109 25 ) ) PUNCT ajst-31990 109 26 , , PUNCT ajst-31990 109 27 the the DET ajst-31990 109 28 weighted weight VERB ajst-31990 109 29 average average ADJ ajst-31990 109 30 f1 f1 NOUN ajst-31990 109 31 score score NOUN ajst-31990 109 32 ( ( PUNCT ajst-31990 109 33 0.88 0.88 NUM ajst-31990 109 34 ) ) PUNCT ajst-31990 109 35 was be AUX ajst-31990 109 36 higher high ADJ ajst-31990 109 37 than than ADP ajst-31990 109 38 the the DET ajst-31990 109 39 macro macro NOUN ajst-31990 109 40 - - NOUN ajst-31990 109 41 mean mean ADJ ajst-31990 109 42 ( ( PUNCT ajst-31990 109 43 0.85 0.85 NUM ajst-31990 109 44 ) ) PUNCT ajst-31990 109 45 , , PUNCT ajst-31990 109 46 reflecting reflect VERB ajst-31990 109 47 the the DET ajst-31990 109 48 model model NOUN ajst-31990 109 49 's 's PART ajst-31990 109 50 superiority superiority NOUN ajst-31990 109 51 in in ADP ajst-31990 109 52 classifying classify VERB ajst-31990 109 53 the the DET ajst-31990 109 54 8 8 NUM ajst-31990 109 55 majority majority NOUN ajst-31990 109 56 of of ADP ajst-31990 109 57 categories category NOUN ajst-31990 109 58 ( ( PUNCT ajst-31990 109 59 1.0 1.0 NUM ajst-31990 109 60 ) ) PUNCT ajst-31990 109 61 . . PUNCT ajst-31990 110 1 despite despite SCONJ ajst-31990 110 2 the the DET ajst-31990 110 3 overall overall ADJ ajst-31990 110 4 good good ADJ ajst-31990 110 5 performance performance NOUN ajst-31990 110 6 of of ADP ajst-31990 110 7 the the DET ajst-31990 110 8 model model NOUN ajst-31990 110 9 , , PUNCT ajst-31990 110 10 it it PRON ajst-31990 110 11 is be AUX ajst-31990 110 12 still still ADV ajst-31990 110 13 necessary necessary ADJ ajst-31990 110 14 to to PART ajst-31990 110 15 optimise optimise VERB ajst-31990 110 16 the the DET ajst-31990 110 17 recognition recognition NOUN ajst-31990 110 18 effect effect NOUN ajst-31990 110 19 of of ADP ajst-31990 110 20 a a DET ajst-31990 110 21 few few ADJ ajst-31990 110 22 classes class NOUN ajst-31990 110 23 by by ADP ajst-31990 110 24 adjusting adjust VERB ajst-31990 110 25 the the DET ajst-31990 110 26 classification classification NOUN ajst-31990 110 27 threshold threshold NOUN ajst-31990 110 28 or or CCONJ ajst-31990 110 29 category category NOUN ajst-31990 110 30 balancing balance VERB ajst-31990 110 31 strategy strategy NOUN ajst-31990 110 32 for for ADP ajst-31990 110 33 the the DET ajst-31990 110 34 lack lack NOUN ajst-31990 110 35 of of ADP ajst-31990 110 36 negative negative ADJ ajst-31990 110 37 class class NOUN ajst-31990 110 38 recall recall NOUN ajst-31990 110 39 ability ability NOUN ajst-31990 110 40 . . PUNCT ajst-31990 111 1 table table NOUN ajst-31990 111 2 3 3 NUM ajst-31990 111 3 . . PUNCT ajst-31990 111 4 classification classification NOUN ajst-31990 111 5 report report NOUN ajst-31990 111 6 for for ADP ajst-31990 111 7 logistic logistic ADJ ajst-31990 111 8 regression regression NOUN ajst-31990 111 9 logistic logistic ADJ ajst-31990 111 10 regression regression NOUN ajst-31990 111 11 precision precision NOUN ajst-31990 111 12 recall recall VERB ajst-31990 111 13 f1 f1 NOUN ajst-31990 111 14 - - PUNCT ajst-31990 111 15 score score NOUN ajst-31990 111 16 support support NOUN ajst-31990 111 17 0.0 0.0 NUM ajst-31990 111 18 0.83 0.83 NUM ajst-31990 111 19 0.71 0.71 NUM ajst-31990 111 20 0.77 0.77 NUM ajst-31990 111 21 14 14 NUM ajst-31990 111 22 1.0 1.0 NUM ajst-31990 111 23 0.90 0.90 NUM ajst-31990 111 24 0.95 0.95 NUM ajst-31990 111 25 0.92 0.92 NUM ajst-31990 111 26 37 37 NUM ajst-31990 111 27 accuracy accuracy NOUN ajst-31990 111 28 / / SYM ajst-31990 111 29 / / SYM ajst-31990 111 30 0.88 0.88 NUM ajst-31990 111 31 51 51 NUM ajst-31990 111 32 macro macro NOUN ajst-31990 111 33 avg avg NOUN ajst-31990 111 34 0.87 0.87 NUM ajst-31990 111 35 0.83 0.83 NUM ajst-31990 111 36 0.85 0.85 NUM ajst-31990 111 37 51 51 NUM ajst-31990 111 38 weighted weight VERB ajst-31990 111 39 avg avg NOUN ajst-31990 111 40 0.88 0.88 NUM ajst-31990 111 41 0.88 0.88 NUM ajst-31990 111 42 0.88 0.88 NUM ajst-31990 111 43 51 51 NUM ajst-31990 111 44 the the DET ajst-31990 111 45 confusion confusion NOUN ajst-31990 111 46 matrix matrix NOUN ajst-31990 111 47 analysis analysis NOUN ajst-31990 111 48 of of ADP ajst-31990 111 49 the the DET ajst-31990 111 50 svm svm ADJ ajst-31990 111 51 model model NOUN ajst-31990 111 52 is be AUX ajst-31990 111 53 shown show VERB ajst-31990 111 54 in in ADP ajst-31990 111 55 figure figure NOUN ajst-31990 111 56 3 3 NUM ajst-31990 111 57 , , PUNCT ajst-31990 111 58 which which PRON ajst-31990 111 59 shows show VERB ajst-31990 111 60 a a DET ajst-31990 111 61 more more ADV ajst-31990 111 62 balanced balanced ADJ ajst-31990 111 63 classification classification NOUN ajst-31990 111 64 of of ADP ajst-31990 111 65 positive positive ADJ ajst-31990 111 66 ( ( PUNCT ajst-31990 111 67 1.0 1.0 NUM ajst-31990 111 68 ) ) PUNCT ajst-31990 111 69 and and CCONJ ajst-31990 111 70 negative negative ADJ ajst-31990 111 71 ( ( PUNCT ajst-31990 111 72 0.0 0.0 NUM ajst-31990 111 73 ) ) PUNCT ajst-31990 111 74 classes class NOUN ajst-31990 111 75 on on ADP ajst-31990 111 76 the the DET ajst-31990 111 77 validation validation NOUN ajst-31990 111 78 set.the set.the DET ajst-31990 111 79 model model NOUN ajst-31990 111 80 correctly correctly ADV ajst-31990 111 81 identified identify VERB ajst-31990 111 82 24 24 NUM ajst-31990 111 83 positive positive ADJ ajst-31990 111 84 examples example NOUN ajst-31990 111 85 ( ( PUNCT ajst-31990 111 86 tp tp NOUN ajst-31990 111 87 ) ) PUNCT ajst-31990 111 88 and and CCONJ ajst-31990 111 89 9 9 NUM ajst-31990 111 90 negative negative ADJ ajst-31990 111 91 examples example NOUN ajst-31990 111 92 ( ( PUNCT ajst-31990 111 93 tn tn NOUN ajst-31990 111 94 ) ) PUNCT ajst-31990 111 95 , , PUNCT ajst-31990 111 96 but but CCONJ ajst-31990 111 97 there there PRON ajst-31990 111 98 were be VERB ajst-31990 111 99 13 13 NUM ajst-31990 111 100 positive positive ADJ ajst-31990 111 101 examples example NOUN ajst-31990 111 102 misclassified misclassifie VERB ajst-31990 111 103 as as ADP ajst-31990 111 104 negative negative ADJ ajst-31990 111 105 class class NOUN ajst-31990 111 106 ( ( PUNCT ajst-31990 111 107 fn fn NOUN ajst-31990 111 108 ) ) PUNCT ajst-31990 111 109 and and CCONJ ajst-31990 111 110 5 5 NUM ajst-31990 111 111 negative negative ADJ ajst-31990 111 112 examples example NOUN ajst-31990 111 113 misclassified misclassifie VERB ajst-31990 111 114 as as ADP ajst-31990 111 115 positive positive ADJ ajst-31990 111 116 class class NOUN ajst-31990 111 117 ( ( PUNCT ajst-31990 111 118 fp fp NOUN ajst-31990 111 119 ) ) PUNCT ajst-31990 111 120 . . PUNCT ajst-31990 112 1 calculations calculation NOUN ajst-31990 112 2 yield yield VERB ajst-31990 112 3 a a DET ajst-31990 112 4 positive positive ADJ ajst-31990 112 5 category category NOUN ajst-31990 112 6 recall recall NOUN ajst-31990 112 7 rate rate NOUN ajst-31990 112 8 of of ADP ajst-31990 112 9 64.9 64.9 NUM ajst-31990 112 10 % % NOUN ajst-31990 112 11 ( ( PUNCT ajst-31990 112 12 24/37 24/37 NUM ajst-31990 112 13 ) ) PUNCT ajst-31990 112 14 and and CCONJ ajst-31990 112 15 a a DET ajst-31990 112 16 negative negative ADJ ajst-31990 112 17 category category NOUN ajst-31990 112 18 precision precision NOUN ajst-31990 112 19 rate rate NOUN ajst-31990 112 20 of of ADP ajst-31990 112 21 64.3 64.3 NUM ajst-31990 112 22 % % NOUN ajst-31990 112 23 ( ( PUNCT ajst-31990 112 24 9/14 9/14 NOUN ajst-31990 112 25 ) ) PUNCT ajst-31990 112 26 , , PUNCT ajst-31990 112 27 indicating indicate VERB ajst-31990 112 28 that that SCONJ ajst-31990 112 29 the the DET ajst-31990 112 30 model model NOUN ajst-31990 112 31 has have VERB ajst-31990 112 32 similar similar ADJ ajst-31990 112 33 discriminative discriminative NOUN ajst-31990 112 34 ability ability NOUN ajst-31990 112 35 for for ADP ajst-31990 112 36 the the DET ajst-31990 112 37 two two NUM ajst-31990 112 38 categories category NOUN ajst-31990 112 39 , , PUNCT ajst-31990 112 40 but but CCONJ ajst-31990 112 41 there there PRON ajst-31990 112 42 is be VERB ajst-31990 112 43 still still ADV ajst-31990 112 44 room room NOUN ajst-31990 112 45 for for ADP ajst-31990 112 46 improvement improvement NOUN ajst-31990 112 47 in in ADP ajst-31990 112 48 the the DET ajst-31990 112 49 overall overall ADJ ajst-31990 112 50 recognition recognition NOUN ajst-31990 112 51 effect effect NOUN ajst-31990 112 52 , , PUNCT ajst-31990 112 53 especially especially ADV ajst-31990 112 54 in in ADP ajst-31990 112 55 reducing reduce VERB ajst-31990 112 56 the the DET ajst-31990 112 57 omission omission NOUN ajst-31990 112 58 rate rate NOUN ajst-31990 112 59 of of ADP ajst-31990 112 60 positive positive ADJ ajst-31990 112 61 category category NOUN ajst-31990 112 62 samples sample NOUN ajst-31990 112 63 . . PUNCT ajst-31990 113 1 figure figure NOUN ajst-31990 113 2 3 3 NUM ajst-31990 113 3 . . PUNCT ajst-31990 113 4 classification classification NOUN ajst-31990 113 5 report report NOUN ajst-31990 113 6 for for ADP ajst-31990 113 7 logistic logistic ADJ ajst-31990 113 8 regression regression NOUN ajst-31990 113 9 ( ( PUNCT ajst-31990 113 10 picture picture NOUN ajst-31990 113 11 credit credit NOUN ajst-31990 113 12 : : PUNCT ajst-31990 113 13 original original ADJ ajst-31990 113 14 ) ) PUNCT ajst-31990 113 15 the the DET ajst-31990 113 16 classification classification NOUN ajst-31990 113 17 results result NOUN ajst-31990 113 18 of of ADP ajst-31990 113 19 the the DET ajst-31990 113 20 svm svm ADJ ajst-31990 113 21 model model NOUN ajst-31990 113 22 are be AUX ajst-31990 113 23 shown show VERB ajst-31990 113 24 in in ADP ajst-31990 113 25 table table NOUN ajst-31990 113 26 4 4 NUM ajst-31990 113 27 with with ADP ajst-31990 113 28 an an DET ajst-31990 113 29 overall overall ADJ ajst-31990 113 30 accuracy accuracy NOUN ajst-31990 113 31 of of ADP ajst-31990 113 32 65 65 NUM ajst-31990 113 33 % % NOUN ajst-31990 113 34 on on ADP ajst-31990 113 35 the the DET ajst-31990 113 36 test test NOUN ajst-31990 113 37 set set NOUN ajst-31990 113 38 ( ( PUNCT ajst-31990 113 39 n=51 n=51 NUM ajst-31990 113 40 ) ) PUNCT ajst-31990 113 41 . . PUNCT ajst-31990 114 1 in in ADP ajst-31990 114 2 terms term NOUN ajst-31990 114 3 of of ADP ajst-31990 114 4 category category NOUN ajst-31990 114 5 performance performance NOUN ajst-31990 114 6 , , PUNCT ajst-31990 114 7 the the DET ajst-31990 114 8 model model NOUN ajst-31990 114 9 outperforms outperform VERB ajst-31990 114 10 the the DET ajst-31990 114 11 negative negative ADJ ajst-31990 114 12 category category NOUN ajst-31990 114 13 ( ( PUNCT ajst-31990 114 14 0.0 0.0 NUM ajst-31990 114 15 ) ) PUNCT ajst-31990 114 16 for for ADP ajst-31990 114 17 the the DET ajst-31990 114 18 positive positive ADJ ajst-31990 114 19 category category NOUN ajst-31990 114 20 ( ( PUNCT ajst-31990 114 21 1.0 1.0 NUM ajst-31990 114 22 ) ) PUNCT ajst-31990 114 23 , , PUNCT ajst-31990 114 24 with with ADP ajst-31990 114 25 an an DET ajst-31990 114 26 accuracy accuracy NOUN ajst-31990 114 27 of of ADP ajst-31990 114 28 83 83 NUM ajst-31990 114 29 % % NOUN ajst-31990 114 30 , , PUNCT ajst-31990 114 31 while while SCONJ ajst-31990 114 32 the the DET ajst-31990 114 33 negative negative ADJ ajst-31990 114 34 category category NOUN ajst-31990 114 35 has have VERB ajst-31990 114 36 an an DET ajst-31990 114 37 accuracy accuracy NOUN ajst-31990 114 38 of of ADP ajst-31990 114 39 only only ADV ajst-31990 114 40 41 41 NUM ajst-31990 114 41 % % NOUN ajst-31990 114 42 . . PUNCT ajst-31990 115 1 in in ADP ajst-31990 115 2 terms term NOUN ajst-31990 115 3 of of ADP ajst-31990 115 4 recall recall NOUN ajst-31990 115 5 , , PUNCT ajst-31990 115 6 the the DET ajst-31990 115 7 negative negative ADJ ajst-31990 115 8 category category NOUN ajst-31990 115 9 ( ( PUNCT ajst-31990 115 10 64 64 NUM ajst-31990 115 11 % % NOUN ajst-31990 115 12 ) ) PUNCT ajst-31990 115 13 is be AUX ajst-31990 115 14 slightly slightly ADV ajst-31990 115 15 higher high ADJ ajst-31990 115 16 than than ADP ajst-31990 115 17 the the DET ajst-31990 115 18 positive positive ADJ ajst-31990 115 19 category category NOUN ajst-31990 115 20 ( ( PUNCT ajst-31990 115 21 65 65 NUM ajst-31990 115 22 % % NOUN ajst-31990 115 23 ) ) PUNCT ajst-31990 115 24 , , PUNCT ajst-31990 115 25 indicating indicate VERB ajst-31990 115 26 that that SCONJ ajst-31990 115 27 the the DET ajst-31990 115 28 model model NOUN ajst-31990 115 29 is be AUX ajst-31990 115 30 relatively relatively ADV ajst-31990 115 31 sensitive sensitive ADJ ajst-31990 115 32 to to ADP ajst-31990 115 33 the the DET ajst-31990 115 34 identification identification NOUN ajst-31990 115 35 of of ADP ajst-31990 115 36 negative negative ADJ ajst-31990 115 37 samples sample NOUN ajst-31990 115 38 , , PUNCT ajst-31990 115 39 but but CCONJ ajst-31990 115 40 has have VERB ajst-31990 115 41 a a DET ajst-31990 115 42 higher high ADJ ajst-31990 115 43 false false ADJ ajst-31990 115 44 positive positive ADJ ajst-31990 115 45 rate rate NOUN ajst-31990 115 46 . . PUNCT ajst-31990 116 1 this this DET ajst-31990 116 2 difference difference NOUN ajst-31990 116 3 is be AUX ajst-31990 116 4 further far ADV ajst-31990 116 5 confirmed confirm VERB ajst-31990 116 6 by by ADP ajst-31990 116 7 the the DET ajst-31990 116 8 f1 f1 PROPN ajst-31990 116 9 scores score NOUN ajst-31990 116 10 , , PUNCT ajst-31990 116 11 which which PRON ajst-31990 116 12 are be AUX ajst-31990 116 13 0.73 0.73 NUM ajst-31990 116 14 for for ADP ajst-31990 116 15 the the DET ajst-31990 116 16 positive positive ADJ ajst-31990 116 17 class class NOUN ajst-31990 116 18 and and CCONJ ajst-31990 116 19 0.50 0.50 NUM ajst-31990 116 20 for for ADP ajst-31990 116 21 the the DET ajst-31990 116 22 negative negative ADJ ajst-31990 116 23 class class NOUN ajst-31990 116 24 . . PUNCT ajst-31990 117 1 it it PRON ajst-31990 117 2 is be AUX ajst-31990 117 3 noteworthy noteworthy ADJ ajst-31990 117 4 that that SCONJ ajst-31990 117 5 the the DET ajst-31990 117 6 weighted weight VERB ajst-31990 117 7 average average ADJ ajst-31990 117 8 indicator indicator NOUN ajst-31990 117 9 ( ( PUNCT ajst-31990 117 10 f1 f1 NOUN ajst-31990 117 11 = = NOUN ajst-31990 117 12 0.66 0.66 NUM ajst-31990 117 13 ) ) PUNCT ajst-31990 117 14 outperforms outperform VERB ajst-31990 117 15 the the DET ajst-31990 117 16 macro macro NOUN ajst-31990 117 17 - - NOUN ajst-31990 117 18 mean mean ADJ ajst-31990 117 19 ( ( PUNCT ajst-31990 117 20 f1 f1 NOUN ajst-31990 117 21 = = SYM ajst-31990 117 22 0.61 0.61 NUM ajst-31990 117 23 ) ) PUNCT ajst-31990 117 24 , , PUNCT ajst-31990 117 25 reflecting reflect VERB ajst-31990 117 26 the the DET ajst-31990 117 27 model model NOUN ajst-31990 117 28 's 's PART ajst-31990 117 29 dominance dominance NOUN ajst-31990 117 30 over over ADP ajst-31990 117 31 the the DET ajst-31990 117 32 majority majority NOUN ajst-31990 117 33 class class NOUN ajst-31990 117 34 ( ( PUNCT ajst-31990 117 35 1.0 1.0 NUM ajst-31990 117 36 ) ) PUNCT ajst-31990 117 37 . . PUNCT ajst-31990 118 1 the the DET ajst-31990 118 2 results result NOUN ajst-31990 118 3 show show VERB ajst-31990 118 4 that that SCONJ ajst-31990 118 5 although although SCONJ ajst-31990 118 6 svm svm NOUN ajst-31990 118 7 is be AUX ajst-31990 118 8 more more ADV ajst-31990 118 9 accurate accurate ADJ ajst-31990 118 10 in in ADP ajst-31990 118 11 discriminating discriminate VERB ajst-31990 118 12 positive positive ADJ ajst-31990 118 13 classes class NOUN ajst-31990 118 14 , , PUNCT ajst-31990 118 15 the the DET ajst-31990 118 16 classification classification NOUN ajst-31990 118 17 performance performance NOUN ajst-31990 118 18 for for ADP ajst-31990 118 19 negative negative ADJ ajst-31990 118 20 classes class NOUN ajst-31990 118 21 needs need VERB ajst-31990 118 22 to to PART ajst-31990 118 23 be be AUX ajst-31990 118 24 improved improve VERB ajst-31990 118 25 , , PUNCT ajst-31990 118 26 especially especially ADV ajst-31990 118 27 in in ADP ajst-31990 118 28 reducing reduce VERB ajst-31990 118 29 the the DET ajst-31990 118 30 false false ADJ ajst-31990 118 31 positive positive ADJ ajst-31990 118 32 rate rate NOUN ajst-31990 118 33 . . PUNCT ajst-31990 119 1 the the DET ajst-31990 119 2 overall overall ADJ ajst-31990 119 3 performance performance NOUN ajst-31990 119 4 of of ADP ajst-31990 119 5 the the DET ajst-31990 119 6 model model NOUN ajst-31990 119 7 can can AUX ajst-31990 119 8 be be AUX ajst-31990 119 9 subsequently subsequently ADV ajst-31990 119 10 improved improve VERB ajst-31990 119 11 by by ADP ajst-31990 119 12 adjusting adjust VERB ajst-31990 119 13 the the DET ajst-31990 119 14 category category NOUN ajst-31990 119 15 weights weight NOUN ajst-31990 119 16 or or CCONJ ajst-31990 119 17 optimising optimise VERB ajst-31990 119 18 the the DET ajst-31990 119 19 kernel kernel PROPN ajst-31990 119 20 function function NOUN ajst-31990 119 21 . . PUNCT ajst-31990 120 1 9 9 NUM ajst-31990 120 2 table table NOUN ajst-31990 120 3 4 4 NUM ajst-31990 120 4 . . PUNCT ajst-31990 120 5 classification classification NOUN ajst-31990 120 6 report report NOUN ajst-31990 120 7 for for ADP ajst-31990 120 8 svm svm ADJ ajst-31990 120 9 svm svm PROPN ajst-31990 120 10 precision precision PROPN ajst-31990 120 11 recall recall VERB ajst-31990 120 12 f1 f1 NOUN ajst-31990 120 13 - - PUNCT ajst-31990 120 14 score score NOUN ajst-31990 120 15 support support NOUN ajst-31990 120 16 0.0 0.0 NUM ajst-31990 120 17 0.41 0.41 NUM ajst-31990 120 18 0.64 0.64 NUM ajst-31990 120 19 0.50 0.50 NUM ajst-31990 120 20 14 14 NUM ajst-31990 120 21 1.0 1.0 NUM ajst-31990 120 22 0.83 0.83 NUM ajst-31990 120 23 0.65 0.65 NUM ajst-31990 120 24 0.73 0.73 NUM ajst-31990 120 25 37 37 NUM ajst-31990 120 26 accuracy accuracy NOUN ajst-31990 120 27 / / SYM ajst-31990 120 28 / / SYM ajst-31990 120 29 0.65 0.65 NUM ajst-31990 120 30 51 51 NUM ajst-31990 120 31 macro macro NOUN ajst-31990 120 32 avg avg NOUN ajst-31990 120 33 0.62 0.62 NUM ajst-31990 120 34 0.65 0.65 NUM ajst-31990 120 35 0.61 0.61 NUM ajst-31990 120 36 51 51 NUM ajst-31990 120 37 weighted weight VERB ajst-31990 120 38 avg avg NOUN ajst-31990 120 39 0.71 0.71 NUM ajst-31990 120 40 0.65 0.65 NUM ajst-31990 120 41 0.66 0.66 NUM ajst-31990 120 42 51 51 NUM ajst-31990 120 43 figure figure NOUN ajst-31990 120 44 4 4 NUM ajst-31990 120 45 shows show VERB ajst-31990 120 46 a a DET ajst-31990 120 47 visual visual ADJ ajst-31990 120 48 comparison comparison NOUN ajst-31990 120 49 of of ADP ajst-31990 120 50 the the DET ajst-31990 120 51 overall overall ADJ ajst-31990 120 52 performance performance NOUN ajst-31990 120 53 of of ADP ajst-31990 120 54 the the DET ajst-31990 120 55 two two NUM ajst-31990 120 56 models model NOUN ajst-31990 120 57 , , PUNCT ajst-31990 120 58 the the DET ajst-31990 120 59 logistic logistic ADJ ajst-31990 120 60 regression regression NOUN ajst-31990 120 61 model model NOUN ajst-31990 120 62 is be AUX ajst-31990 120 63 far far ADV ajst-31990 120 64 superior superior ADJ ajst-31990 120 65 to to ADP ajst-31990 120 66 the the DET ajst-31990 120 67 svm svm ADJ ajst-31990 120 68 model model NOUN ajst-31990 120 69 in in ADP ajst-31990 120 70 all all DET ajst-31990 120 71 the the DET ajst-31990 120 72 three three NUM ajst-31990 120 73 main main ADJ ajst-31990 120 74 metrics metric NOUN ajst-31990 120 75 , , PUNCT ajst-31990 120 76 which which PRON ajst-31990 120 77 shows show VERB ajst-31990 120 78 that that SCONJ ajst-31990 120 79 the the DET ajst-31990 120 80 logistic logistic ADJ ajst-31990 120 81 regression regression NOUN ajst-31990 120 82 model model NOUN ajst-31990 120 83 has have VERB ajst-31990 120 84 a a DET ajst-31990 120 85 more more ADV ajst-31990 120 86 balanced balanced ADJ ajst-31990 120 87 performance performance NOUN ajst-31990 120 88 and and CCONJ ajst-31990 120 89 performs perform VERB ajst-31990 120 90 better well ADV ajst-31990 120 91 on on ADP ajst-31990 120 92 most most ADJ ajst-31990 120 93 of of ADP ajst-31990 120 94 the the DET ajst-31990 120 95 classes class NOUN ajst-31990 120 96 . . PUNCT ajst-31990 121 1 figure figure VERB ajst-31990 121 2 4 4 NUM ajst-31990 121 3 . . NOUN ajst-31990 121 4 comparison comparison NOUN ajst-31990 121 5 of of ADP ajst-31990 121 6 the the DET ajst-31990 121 7 overall overall ADJ ajst-31990 121 8 performance performance NOUN ajst-31990 121 9 of of ADP ajst-31990 121 10 the the DET ajst-31990 121 11 two two NUM ajst-31990 121 12 models model NOUN ajst-31990 121 13 ( ( PUNCT ajst-31990 121 14 picture picture NOUN ajst-31990 121 15 credit credit NOUN ajst-31990 121 16 : : PUNCT ajst-31990 121 17 original original ADJ ajst-31990 121 18 ) ) PUNCT ajst-31990 121 19 figure figure NOUN ajst-31990 121 20 5 5 NUM ajst-31990 121 21 shows show VERB ajst-31990 121 22 a a DET ajst-31990 121 23 comparison comparison NOUN ajst-31990 121 24 of of ADP ajst-31990 121 25 the the DET ajst-31990 121 26 patient patient ADJ ajst-31990 121 27 classification classification NOUN ajst-31990 121 28 performance performance NOUN ajst-31990 121 29 of of ADP ajst-31990 121 30 the the DET ajst-31990 121 31 two two NUM ajst-31990 121 32 models model NOUN ajst-31990 121 33 , , PUNCT ajst-31990 121 34 which which PRON ajst-31990 121 35 shows show VERB ajst-31990 121 36 that that SCONJ ajst-31990 121 37 the the DET ajst-31990 121 38 logistic logistic ADJ ajst-31990 121 39 regression regression NOUN ajst-31990 121 40 model model NOUN ajst-31990 121 41 has have VERB ajst-31990 121 42 a a DET ajst-31990 121 43 lower low ADJ ajst-31990 121 44 percentage percentage NOUN ajst-31990 121 45 of of ADP ajst-31990 121 46 misdiagnosed misdiagnose VERB ajst-31990 121 47 patients patient NOUN ajst-31990 121 48 . . PUNCT ajst-31990 122 1 figure figure VERB ajst-31990 122 2 5 5 NUM ajst-31990 122 3 . . PUNCT ajst-31990 123 1 patient patient ADJ ajst-31990 123 2 classification classification NOUN ajst-31990 123 3 performance performance NOUN ajst-31990 123 4 comparison comparison NOUN ajst-31990 123 5 ( ( PUNCT ajst-31990 123 6 picture picture NOUN ajst-31990 123 7 credit credit NOUN ajst-31990 123 8 : : PUNCT ajst-31990 123 9 original original ADJ ajst-31990 123 10 ) ) PUNCT ajst-31990 123 11 0.00 0.00 NUM ajst-31990 123 12 0.20 0.20 NUM ajst-31990 123 13 0.40 0.40 NUM ajst-31990 123 14 0.60 0.60 NUM ajst-31990 123 15 0.80 0.80 NUM ajst-31990 123 16 1.00 1.00 NUM ajst-31990 123 17 accuracy accuracy NOUN ajst-31990 123 18 macro macro ADJ ajst-31990 123 19 avg avg PROPN ajst-31990 123 20 f1 f1 PROPN ajst-31990 123 21 weighted weight VERB ajst-31990 123 22 avg avg PROPN ajst-31990 123 23 f1 f1 PROPN ajst-31990 123 24 comparison comparison NOUN ajst-31990 123 25 of of ADP ajst-31990 123 26 overall overall ADJ ajst-31990 123 27 model model NOUN ajst-31990 123 28 performance performance NOUN ajst-31990 123 29 logistic logistic ADJ ajst-31990 123 30 regression regression NOUN ajst-31990 123 31 svm svm VERB ajst-31990 123 32 8 8 NUM ajst-31990 123 33 3 3 NUM ajst-31990 123 34 % % NOUN ajst-31990 123 35 7 7 NUM ajst-31990 123 36 1 1 NUM ajst-31990 123 37 % % NOUN ajst-31990 123 38 7 7 NUM ajst-31990 123 39 7 7 NUM ajst-31990 123 40 % % NOUN ajst-31990 123 41 4 4 NUM ajst-31990 123 42 1 1 NUM ajst-31990 123 43 % % NOUN ajst-31990 123 44 6 6 NUM ajst-31990 123 45 4 4 NUM ajst-31990 123 46 % % NOUN ajst-31990 123 47 5 5 NUM ajst-31990 123 48 0 0 NUM ajst-31990 123 49 % % NOUN ajst-31990 123 50 p p NOUN ajst-31990 123 51 r r NOUN ajst-31990 123 52 e e NOUN ajst-31990 123 53 c c NOUN ajst-31990 124 1 i i PRON ajst-31990 124 2 s s VERB ajst-31990 124 3 i i PRON ajst-31990 124 4 o o NOUN ajst-31990 125 1 n n CCONJ ajst-31990 125 2 r r NOUN ajst-31990 125 3 e e NOUN ajst-31990 125 4 c c NOUN ajst-31990 125 5 a a DET ajst-31990 125 6 l l NOUN ajst-31990 125 7 l l NOUN ajst-31990 125 8 f f X ajst-31990 126 1 1 1 NUM ajst-31990 126 2 s s NOUN ajst-31990 126 3 c c NOUN ajst-31990 126 4 o o NOUN ajst-31990 126 5 r r NOUN ajst-31990 126 6 e e NOUN ajst-31990 126 7 category category NOUN ajst-31990 126 8 0 0 NUM ajst-31990 127 1 ( ( PUNCT ajst-31990 127 2 patient patient NOUN ajst-31990 127 3 ) ) PUNCT ajst-31990 127 4 logistic logistic ADJ ajst-31990 127 5 regression regression NOUN ajst-31990 127 6 svm svm VERB ajst-31990 127 7 10 10 NUM ajst-31990 127 8 figure figure NOUN ajst-31990 127 9 6 6 NUM ajst-31990 127 10 shows show VERB ajst-31990 127 11 the the DET ajst-31990 127 12 comparison comparison NOUN ajst-31990 127 13 of of ADP ajst-31990 127 14 the the DET ajst-31990 127 15 classification classification NOUN ajst-31990 127 16 performance performance NOUN ajst-31990 127 17 of of ADP ajst-31990 127 18 the the DET ajst-31990 127 19 two two NUM ajst-31990 127 20 models model NOUN ajst-31990 127 21 for for ADP ajst-31990 127 22 healthy healthy ADJ ajst-31990 127 23 people people NOUN ajst-31990 127 24 , , PUNCT ajst-31990 127 25 which which PRON ajst-31990 127 26 shows show VERB ajst-31990 127 27 that that SCONJ ajst-31990 127 28 the the DET ajst-31990 127 29 logistic logistic ADJ ajst-31990 127 30 regression regression NOUN ajst-31990 127 31 model model NOUN ajst-31990 127 32 has have VERB ajst-31990 127 33 a a DET ajst-31990 127 34 lower low ADJ ajst-31990 127 35 percentage percentage NOUN ajst-31990 127 36 of of ADP ajst-31990 127 37 misdiagnosed misdiagnose VERB ajst-31990 127 38 healthy healthy ADJ ajst-31990 127 39 people people NOUN ajst-31990 127 40 and and CCONJ ajst-31990 127 41 is be AUX ajst-31990 127 42 more more ADV ajst-31990 127 43 reliable reliable ADJ ajst-31990 127 44 in in ADP ajst-31990 127 45 identifying identify VERB ajst-31990 127 46 healthy healthy ADJ ajst-31990 127 47 people people NOUN ajst-31990 127 48 . . PUNCT ajst-31990 128 1 figure figure VERB ajst-31990 128 2 6 6 NUM ajst-31990 128 3 . . PUNCT ajst-31990 129 1 healthy healthy ADJ ajst-31990 129 2 people people NOUN ajst-31990 129 3 classification classification NOUN ajst-31990 129 4 performance performance NOUN ajst-31990 129 5 comparison comparison NOUN ajst-31990 129 6 ( ( PUNCT ajst-31990 129 7 picture picture NOUN ajst-31990 129 8 credit credit NOUN ajst-31990 129 9 : : PUNCT ajst-31990 129 10 original original ADJ ajst-31990 129 11 ) ) PUNCT ajst-31990 129 12 2.9 2.9 NUM ajst-31990 129 13 . . PUNCT ajst-31990 130 1 comprehensive comprehensive ADJ ajst-31990 130 2 performance performance NOUN ajst-31990 130 3 analysis analysis NOUN ajst-31990 130 4 logistic logistic ADJ ajst-31990 130 5 regression regression NOUN ajst-31990 130 6 , , PUNCT ajst-31990 130 7 xgboost xgboost ADV ajst-31990 130 8 , , PUNCT ajst-31990 130 9 and and CCONJ ajst-31990 130 10 svm svm PROPN ajst-31990 130 11 were be AUX ajst-31990 130 12 each each PRON ajst-31990 130 13 used use VERB ajst-31990 130 14 in in ADP ajst-31990 130 15 a a DET ajst-31990 130 16 pd pd NOUN ajst-31990 130 17 prediction prediction NOUN ajst-31990 130 18 task task NOUN ajst-31990 130 19 . . PUNCT ajst-31990 131 1 among among ADP ajst-31990 131 2 them they PRON ajst-31990 131 3 , , PUNCT ajst-31990 131 4 the the DET ajst-31990 131 5 logistic logistic ADJ ajst-31990 131 6 regression regression NOUN ajst-31990 131 7 validation validation NOUN ajst-31990 131 8 set set NOUN ajst-31990 131 9 has have VERB ajst-31990 131 10 an an DET ajst-31990 131 11 accuracy accuracy NOUN ajst-31990 131 12 of of ADP ajst-31990 131 13 81.47 81.47 NUM ajst-31990 131 14 % % NOUN ajst-31990 131 15 , , PUNCT ajst-31990 131 16 the the DET ajst-31990 131 17 recall recall NOUN ajst-31990 131 18 rate rate NOUN ajst-31990 131 19 of of ADP ajst-31990 131 20 healthy healthy ADJ ajst-31990 131 21 people people NOUN ajst-31990 131 22 is be AUX ajst-31990 131 23 95 95 NUM ajst-31990 131 24 % % NOUN ajst-31990 131 25 while while SCONJ ajst-31990 131 26 the the DET ajst-31990 131 27 recall recall NOUN ajst-31990 131 28 rate rate NOUN ajst-31990 131 29 of of ADP ajst-31990 131 30 patients patient NOUN ajst-31990 131 31 is be AUX ajst-31990 131 32 71 71 NUM ajst-31990 131 33 % % NOUN ajst-31990 131 34 , , PUNCT ajst-31990 131 35 and and CCONJ ajst-31990 131 36 the the DET ajst-31990 131 37 roc roc PROPN ajst-31990 131 38 auc auc NOUN ajst-31990 131 39 value value NOUN ajst-31990 131 40 of of ADP ajst-31990 131 41 0.89 0.89 NUM ajst-31990 131 42 shows show VERB ajst-31990 131 43 that that SCONJ ajst-31990 131 44 it it PRON ajst-31990 131 45 has have VERB ajst-31990 131 46 excellent excellent ADJ ajst-31990 131 47 ability ability NOUN ajst-31990 131 48 to to PART ajst-31990 131 49 distinguish distinguish VERB ajst-31990 131 50 between between ADP ajst-31990 131 51 positive positive ADJ ajst-31990 131 52 and and CCONJ ajst-31990 131 53 negative negative ADJ ajst-31990 131 54 classes class NOUN ajst-31990 131 55 , , PUNCT ajst-31990 131 56 and and CCONJ ajst-31990 131 57 it it PRON ajst-31990 131 58 is be AUX ajst-31990 131 59 stable stable ADJ ajst-31990 131 60 and and CCONJ ajst-31990 131 61 computationally computationally ADV ajst-31990 131 62 efficient efficient ADJ ajst-31990 131 63 in in ADP ajst-31990 131 64 recognising recognise VERB ajst-31990 131 65 the the DET ajst-31990 131 66 majority majority NOUN ajst-31990 131 67 of of ADP ajst-31990 131 68 the the DET ajst-31990 131 69 classes class NOUN ajst-31990 131 70 ( ( PUNCT ajst-31990 131 71 healthy healthy ADJ ajst-31990 131 72 people people NOUN ajst-31990 131 73 ) ) PUNCT ajst-31990 131 74 after after ADP ajst-31990 131 75 oversampling oversample VERB ajst-31990 131 76 . . PUNCT ajst-31990 132 1 the the DET ajst-31990 132 2 xgboost xgboost PROPN ajst-31990 132 3 validation validation PROPN ajst-31990 132 4 set set NOUN ajst-31990 132 5 had have VERB ajst-31990 132 6 an an DET ajst-31990 132 7 accuracy accuracy NOUN ajst-31990 132 8 of of ADP ajst-31990 132 9 78.2 78.2 NUM ajst-31990 132 10 % % NOUN ajst-31990 132 11 and and CCONJ ajst-31990 132 12 a a DET ajst-31990 132 13 patient patient ADJ ajst-31990 132 14 recall recall NOUN ajst-31990 132 15 rate rate NOUN ajst-31990 132 16 of of ADP ajst-31990 132 17 only only ADV ajst-31990 132 18 65 65 NUM ajst-31990 132 19 % % NOUN ajst-31990 132 20 , , PUNCT ajst-31990 132 21 with with ADP ajst-31990 132 22 a a DET ajst-31990 132 23 tendency tendency NOUN ajst-31990 132 24 to to ADP ajst-31990 132 25 overfitting overfitte VERB ajst-31990 132 26 ( ( PUNCT ajst-31990 132 27 auc auc NOUN ajst-31990 132 28 of of ADP ajst-31990 132 29 0.95 0.95 NUM ajst-31990 132 30 for for ADP ajst-31990 132 31 the the DET ajst-31990 132 32 training training NOUN ajst-31990 132 33 set set NOUN ajst-31990 132 34 and and CCONJ ajst-31990 132 35 0.82 0.82 NUM ajst-31990 132 36 for for ADP ajst-31990 132 37 the the DET ajst-31990 132 38 validation validation NOUN ajst-31990 132 39 set set NOUN ajst-31990 132 40 ) ) PUNCT ajst-31990 132 41 , , PUNCT ajst-31990 132 42 showing show VERB ajst-31990 132 43 a a DET ajst-31990 132 44 lack lack NOUN ajst-31990 132 45 of of ADP ajst-31990 132 46 ability ability NOUN ajst-31990 132 47 to to PART ajst-31990 132 48 identify identify VERB ajst-31990 132 49 a a DET ajst-31990 132 50 small small ADJ ajst-31990 132 51 number number NOUN ajst-31990 132 52 of of ADP ajst-31990 132 53 classes class NOUN ajst-31990 132 54 . . PUNCT ajst-31990 133 1 the the DET ajst-31990 133 2 svm svm PROPN ajst-31990 133 3 validation validation NOUN ajst-31990 133 4 set set NOUN ajst-31990 133 5 had have VERB ajst-31990 133 6 an an DET ajst-31990 133 7 accuracy accuracy NOUN ajst-31990 133 8 of of ADP ajst-31990 133 9 75.6 75.6 NUM ajst-31990 133 10 % % NOUN ajst-31990 133 11 , , PUNCT ajst-31990 133 12 a a DET ajst-31990 133 13 patient patient ADJ ajst-31990 133 14 recall recall NOUN ajst-31990 133 15 rate rate NOUN ajst-31990 133 16 of of ADP ajst-31990 133 17 60 60 NUM ajst-31990 133 18 % % NOUN ajst-31990 133 19 , , PUNCT ajst-31990 133 20 and and CCONJ ajst-31990 133 21 an an DET ajst-31990 133 22 roc roc NOUN ajst-31990 133 23 auc auc NOUN ajst-31990 133 24 of of ADP ajst-31990 133 25 0.76 0.76 NUM ajst-31990 133 26 , , PUNCT ajst-31990 133 27 resulting result VERB ajst-31990 133 28 in in ADP ajst-31990 133 29 an an DET ajst-31990 133 30 overall overall ADJ ajst-31990 133 31 poor poor ADJ ajst-31990 133 32 performance performance NOUN ajst-31990 133 33 . . PUNCT ajst-31990 134 1 the the DET ajst-31990 134 2 dataset dataset NOUN ajst-31990 134 3 was be AUX ajst-31990 134 4 approximately approximately ADV ajst-31990 134 5 72.5 72.5 NUM ajst-31990 134 6 % % NOUN ajst-31990 134 7 ( ( PUNCT ajst-31990 134 8 37/51 37/51 NUM ajst-31990 134 9 ) ) PUNCT ajst-31990 134 10 healthy healthy ADJ ajst-31990 134 11 , , PUNCT ajst-31990 134 12 making make VERB ajst-31990 134 13 the the DET ajst-31990 134 14 model model NOUN ajst-31990 134 15 predisposed predispose VERB ajst-31990 134 16 to to PART ajst-31990 134 17 predict predict VERB ajst-31990 134 18 the the DET ajst-31990 134 19 majority majority NOUN ajst-31990 134 20 class class NOUN ajst-31990 134 21 . . PUNCT ajst-31990 135 1 logistic logistic ADJ ajst-31990 135 2 regression regression NOUN ajst-31990 135 3 over over ADP ajst-31990 135 4 - - PUNCT ajst-31990 135 5 sampling sample VERB ajst-31990 135 6 balanced balanced ADJ ajst-31990 135 7 data datum NOUN ajst-31990 135 8 resulted result VERB ajst-31990 135 9 in in ADP ajst-31990 135 10 the the DET ajst-31990 135 11 best good ADJ ajst-31990 135 12 overall overall ADJ ajst-31990 135 13 performance performance NOUN ajst-31990 135 14 and and CCONJ ajst-31990 135 15 excelled excel VERB ajst-31990 135 16 in in ADP ajst-31990 135 17 reducing reduce VERB ajst-31990 135 18 misdiagnosis misdiagnosis NOUN ajst-31990 135 19 in in ADP ajst-31990 135 20 healthy healthy ADJ ajst-31990 135 21 people people NOUN ajst-31990 135 22 , , PUNCT ajst-31990 135 23 but but CCONJ ajst-31990 135 24 there there PRON ajst-31990 135 25 is be VERB ajst-31990 135 26 still still ADV ajst-31990 135 27 room room NOUN ajst-31990 135 28 for for ADP ajst-31990 135 29 improvement improvement NOUN ajst-31990 135 30 in in ADP ajst-31990 135 31 patient patient ADJ ajst-31990 135 32 recall recall NOUN ajst-31990 135 33 . . PUNCT ajst-31990 136 1 xgboost xgboost PROPN ajst-31990 136 2 's 's PART ajst-31990 136 3 poor poor ADJ ajst-31990 136 4 recognition recognition NOUN ajst-31990 136 5 of of ADP ajst-31990 136 6 minority minority NOUN ajst-31990 136 7 classes class NOUN ajst-31990 136 8 due due ADJ ajst-31990 136 9 to to ADP ajst-31990 136 10 insufficient insufficient ADJ ajst-31990 136 11 handling handling NOUN ajst-31990 136 12 of of ADP ajst-31990 136 13 category category NOUN ajst-31990 136 14 imbalance imbalance NOUN ajst-31990 136 15 can can AUX ajst-31990 136 16 be be AUX ajst-31990 136 17 optimised optimise VERB ajst-31990 136 18 by by ADP ajst-31990 136 19 regularisation regularisation NOUN ajst-31990 136 20 and and CCONJ ajst-31990 136 21 feature feature NOUN ajst-31990 136 22 reduction reduction NOUN ajst-31990 136 23 . . PUNCT ajst-31990 137 1 svm svm PROPN ajst-31990 137 2 also also ADV ajst-31990 137 3 suffers suffer VERB ajst-31990 137 4 from from ADP ajst-31990 137 5 this this DET ajst-31990 137 6 problem problem NOUN ajst-31990 137 7 and and CCONJ ajst-31990 137 8 can can AUX ajst-31990 137 9 be be AUX ajst-31990 137 10 combined combine VERB ajst-31990 137 11 with with ADP ajst-31990 137 12 feature feature NOUN ajst-31990 137 13 dimensionality dimensionality NOUN ajst-31990 137 14 reduction reduction NOUN ajst-31990 137 15 to to PART ajst-31990 137 16 improve improve VERB ajst-31990 137 17 performance performance NOUN ajst-31990 137 18 , , PUNCT ajst-31990 137 19 and and CCONJ ajst-31990 137 20 both both PRON ajst-31990 137 21 have have VERB ajst-31990 137 22 a a DET ajst-31990 137 23 need need NOUN ajst-31990 137 24 for for ADP ajst-31990 137 25 further further ADJ ajst-31990 137 26 improvement improvement NOUN ajst-31990 137 27 in in ADP ajst-31990 137 28 sensitivity sensitivity NOUN ajst-31990 137 29 . . PUNCT ajst-31990 138 1 4 4 X ajst-31990 138 2 . . X ajst-31990 138 3 conclusion conclusion NOUN ajst-31990 138 4 this this DET ajst-31990 138 5 study study NOUN ajst-31990 138 6 systematically systematically ADV ajst-31990 138 7 compares compare VERB ajst-31990 138 8 the the DET ajst-31990 138 9 performance performance NOUN ajst-31990 138 10 differences difference NOUN ajst-31990 138 11 between between ADP ajst-31990 138 12 logistic logistic ADJ ajst-31990 138 13 regression regression NOUN ajst-31990 138 14 , , PUNCT ajst-31990 138 15 xgboost xgboost ADV ajst-31990 138 16 and and CCONJ ajst-31990 138 17 svm svm VERB ajst-31990 138 18 in in ADP ajst-31990 138 19 speech speech NOUN ajst-31990 138 20 feature feature NOUN ajst-31990 138 21 classification classification NOUN ajst-31990 138 22 for for ADP ajst-31990 138 23 pd pd PROPN ajst-31990 138 24 , , PUNCT ajst-31990 138 25 revealing reveal VERB ajst-31990 138 26 the the DET ajst-31990 138 27 applicability applicability NOUN ajst-31990 138 28 and and CCONJ ajst-31990 138 29 limitations limitation NOUN ajst-31990 138 30 of of ADP ajst-31990 138 31 each each DET ajst-31990 138 32 model model NOUN ajst-31990 138 33 in in ADP ajst-31990 138 34 medical medical ADJ ajst-31990 138 35 scenarios scenario NOUN ajst-31990 138 36 . . PUNCT ajst-31990 139 1 logistic logistic ADJ ajst-31990 139 2 regression regression NOUN ajst-31990 139 3 performs perform VERB ajst-31990 139 4 best good ADJ ajst-31990 139 5 with with ADP ajst-31990 139 6 81.47 81.47 NUM ajst-31990 139 7 % % NOUN ajst-31990 139 8 validation validation NOUN ajst-31990 139 9 accuracy accuracy NOUN ajst-31990 139 10 and and CCONJ ajst-31990 139 11 95 95 NUM ajst-31990 139 12 % % NOUN ajst-31990 139 13 healthy healthy ADJ ajst-31990 139 14 person person NOUN ajst-31990 139 15 recall recall NOUN ajst-31990 139 16 , , PUNCT ajst-31990 139 17 and and CCONJ ajst-31990 139 18 its its PRON ajst-31990 139 19 high high ADJ ajst-31990 139 20 interpretability interpretability NOUN ajst-31990 139 21 and and CCONJ ajst-31990 139 22 computational computational ADJ ajst-31990 139 23 efficiency efficiency NOUN ajst-31990 139 24 are be AUX ajst-31990 139 25 suitable suitable ADJ ajst-31990 139 26 for for ADP ajst-31990 139 27 use use NOUN ajst-31990 139 28 as as ADP ajst-31990 139 29 an an DET ajst-31990 139 30 efficient efficient ADJ ajst-31990 139 31 screening screening NOUN ajst-31990 139 32 tool tool NOUN ajst-31990 139 33 in in ADP ajst-31990 139 34 primary primary ADJ ajst-31990 139 35 care care NOUN ajst-31990 139 36 , , PUNCT ajst-31990 139 37 but but CCONJ ajst-31990 139 38 the the DET ajst-31990 139 39 71 71 NUM ajst-31990 139 40 % % NOUN ajst-31990 139 41 patient patient NOUN ajst-31990 139 42 recall recall NOUN ajst-31990 139 43 needs need VERB ajst-31990 139 44 to to PART ajst-31990 139 45 be be AUX ajst-31990 139 46 improved improve VERB ajst-31990 139 47 by by ADP ajst-31990 139 48 feature feature NOUN ajst-31990 139 49 optimisation optimisation NOUN ajst-31990 139 50 and and CCONJ ajst-31990 139 51 dynamic dynamic ADJ ajst-31990 139 52 threshold threshold NOUN ajst-31990 139 53 adjustment adjustment NOUN ajst-31990 139 54 . . PUNCT ajst-31990 140 1 although although SCONJ ajst-31990 140 2 xgboost xgboost PROPN ajst-31990 140 3 has have VERB ajst-31990 140 4 a a DET ajst-31990 140 5 low low ADJ ajst-31990 140 6 validation validation NOUN ajst-31990 140 7 set set NOUN ajst-31990 140 8 accuracy accuracy NOUN ajst-31990 140 9 ( ( PUNCT ajst-31990 140 10 78.2 78.2 NUM ajst-31990 140 11 % % NOUN ajst-31990 140 12 ) ) PUNCT ajst-31990 140 13 and and CCONJ ajst-31990 140 14 patient patient ADJ ajst-31990 140 15 recall recall NOUN ajst-31990 140 16 ( ( PUNCT ajst-31990 140 17 65 65 NUM ajst-31990 140 18 % % NOUN ajst-31990 140 19 ) ) PUNCT ajst-31990 140 20 due due ADP ajst-31990 140 21 to to ADP ajst-31990 140 22 category category NOUN ajst-31990 140 23 imbalance imbalance NOUN ajst-31990 140 24 and and CCONJ ajst-31990 140 25 a a DET ajst-31990 140 26 21.43 21.43 NUM ajst-31990 140 27 % % NOUN ajst-31990 140 28 risk risk NOUN ajst-31990 140 29 of of ADP ajst-31990 140 30 overfitting overfitte VERB ajst-31990 140 31 , , PUNCT ajst-31990 140 32 its its PRON ajst-31990 140 33 potential potential NOUN ajst-31990 140 34 for for ADP ajst-31990 140 35 complex complex ADJ ajst-31990 140 36 feature feature NOUN ajst-31990 140 37 interactions interaction NOUN ajst-31990 140 38 can can AUX ajst-31990 140 39 be be AUX ajst-31990 140 40 optimised optimise VERB ajst-31990 140 41 by by ADP ajst-31990 140 42 regularisation regularisation NOUN ajst-31990 140 43 and and CCONJ ajst-31990 140 44 multimodal multimodal NOUN ajst-31990 140 45 data datum NOUN ajst-31990 140 46 fusion fusion NOUN ajst-31990 140 47 . . PUNCT ajst-31990 141 1 svm svm PROPN ajst-31990 141 2 is be AUX ajst-31990 141 3 limited limit VERB ajst-31990 141 4 by by ADP ajst-31990 141 5 the the DET ajst-31990 141 6 current current ADJ ajst-31990 141 7 data datum NOUN ajst-31990 141 8 distribution distribution NOUN ajst-31990 141 9 , , PUNCT ajst-31990 141 10 the the DET ajst-31990 141 11 validation validation NOUN ajst-31990 141 12 accuracy accuracy NOUN ajst-31990 141 13 9 9 NUM ajst-31990 141 14 0 0 NUM ajst-31990 141 15 % % NOUN ajst-31990 141 16 9 9 NUM ajst-31990 141 17 5 5 NUM ajst-31990 141 18 % % NOUN ajst-31990 141 19 9 9 NUM ajst-31990 141 20 2 2 NUM ajst-31990 141 21 % % NOUN ajst-31990 141 22 8 8 NUM ajst-31990 141 23 3 3 NUM ajst-31990 141 24 % % NOUN ajst-31990 141 25 6 6 NUM ajst-31990 141 26 5 5 NUM ajst-31990 141 27 % % NOUN ajst-31990 141 28 7 7 NUM ajst-31990 141 29 3 3 NUM ajst-31990 141 30 % % NOUN ajst-31990 141 31 p p NOUN ajst-31990 141 32 r r NOUN ajst-31990 142 1 e e NOUN ajst-31990 142 2 c c NOUN ajst-31990 143 1 i i PRON ajst-31990 143 2 s s VERB ajst-31990 143 3 i i PRON ajst-31990 143 4 o o NOUN ajst-31990 144 1 n n CCONJ ajst-31990 144 2 r r NOUN ajst-31990 144 3 e e NOUN ajst-31990 144 4 c c NOUN ajst-31990 144 5 a a DET ajst-31990 144 6 l l NOUN ajst-31990 144 7 l l NOUN ajst-31990 144 8 f f X ajst-31990 145 1 1 1 NUM ajst-31990 145 2 s s NOUN ajst-31990 145 3 c c NOUN ajst-31990 145 4 o o NOUN ajst-31990 145 5 r r NOUN ajst-31990 145 6 e e NOUN ajst-31990 145 7 category category NOUN ajst-31990 145 8 1 1 NUM ajst-31990 145 9 ( ( PUNCT ajst-31990 145 10 healthy healthy ADJ ajst-31990 145 11 people people NOUN ajst-31990 145 12 ) ) PUNCT ajst-31990 145 13 logistic logistic ADJ ajst-31990 145 14 regression regression NOUN ajst-31990 145 15 svm svm NOUN ajst-31990 145 16 11 11 NUM ajst-31990 145 17 ( ( PUNCT ajst-31990 145 18 75.6 75.6 NUM ajst-31990 145 19 % % NOUN ajst-31990 145 20 ) ) PUNCT ajst-31990 145 21 and and CCONJ ajst-31990 145 22 patient patient ADJ ajst-31990 145 23 recall recall NOUN ajst-31990 145 24 ( ( PUNCT ajst-31990 145 25 60 60 NUM ajst-31990 145 26 % % NOUN ajst-31990 145 27 ) ) PUNCT ajst-31990 145 28 are be AUX ajst-31990 145 29 both both ADV ajst-31990 145 30 low low ADJ ajst-31990 145 31 , , PUNCT ajst-31990 145 32 and and CCONJ ajst-31990 145 33 should should AUX ajst-31990 145 34 be be AUX ajst-31990 145 35 combined combine VERB ajst-31990 145 36 with with ADP ajst-31990 145 37 feature feature NOUN ajst-31990 145 38 dimensionality dimensionality NOUN ajst-31990 145 39 reduction reduction NOUN ajst-31990 145 40 and and CCONJ ajst-31990 145 41 kernel kernel PROPN ajst-31990 145 42 function function NOUN ajst-31990 145 43 improvement improvement NOUN ajst-31990 145 44 to to PART ajst-31990 145 45 improve improve VERB ajst-31990 145 46 stability stability NOUN ajst-31990 145 47 . . PUNCT ajst-31990 146 1 the the DET ajst-31990 146 2 experimental experimental ADJ ajst-31990 146 3 results result NOUN ajst-31990 146 4 highlight highlight VERB ajst-31990 146 5 the the DET ajst-31990 146 6 need need NOUN ajst-31990 146 7 to to PART ajst-31990 146 8 weigh weigh VERB ajst-31990 146 9 scenario scenario NOUN ajst-31990 146 10 requirements requirement NOUN ajst-31990 146 11 for for ADP ajst-31990 146 12 model model NOUN ajst-31990 146 13 selection selection NOUN ajst-31990 146 14 : : PUNCT ajst-31990 146 15 logistic logistic ADJ ajst-31990 146 16 regression regression NOUN ajst-31990 146 17 has have VERB ajst-31990 146 18 significant significant ADJ ajst-31990 146 19 advantages advantage NOUN ajst-31990 146 20 in in ADP ajst-31990 146 21 reducing reduce VERB ajst-31990 146 22 misdiagnosis misdiagnosis NOUN ajst-31990 146 23 ( ( PUNCT ajst-31990 146 24 high high ADJ ajst-31990 146 25 recall recall NOUN ajst-31990 146 26 ) ) PUNCT ajst-31990 146 27 of of ADP ajst-31990 146 28 healthy healthy ADJ ajst-31990 146 29 people people NOUN ajst-31990 146 30 , , PUNCT ajst-31990 146 31 but but CCONJ ajst-31990 146 32 needs need VERB ajst-31990 146 33 to to PART ajst-31990 146 34 be be AUX ajst-31990 146 35 combined combine VERB ajst-31990 146 36 with with ADP ajst-31990 146 37 a a DET ajst-31990 146 38 physician physician NOUN ajst-31990 146 39 review review NOUN ajst-31990 146 40 mechanism mechanism NOUN ajst-31990 146 41 to to PART ajst-31990 146 42 reduce reduce VERB ajst-31990 146 43 the the DET ajst-31990 146 44 risk risk NOUN ajst-31990 146 45 of of ADP ajst-31990 146 46 missed miss VERB ajst-31990 146 47 diagnosis diagnosis NOUN ajst-31990 146 48 ; ; PUNCT ajst-31990 146 49 xgboost xgboost X ajst-31990 146 50 is be AUX ajst-31990 146 51 expected expect VERB ajst-31990 146 52 to to PART ajst-31990 146 53 break break VERB ajst-31990 146 54 through through ADP ajst-31990 146 55 the the DET ajst-31990 146 56 performance performance NOUN ajst-31990 146 57 bottleneck bottleneck NOUN ajst-31990 146 58 after after ADP ajst-31990 146 59 optimisation optimisation NOUN ajst-31990 146 60 of of ADP ajst-31990 146 61 feature feature NOUN ajst-31990 146 62 engineering engineering NOUN ajst-31990 146 63 ; ; PUNCT ajst-31990 146 64 and and CCONJ ajst-31990 146 65 svm svm PROPN ajst-31990 146 66 can can AUX ajst-31990 146 67 be be AUX ajst-31990 146 68 used use VERB ajst-31990 146 69 as as ADP ajst-31990 146 70 an an DET ajst-31990 146 71 auxiliary auxiliary ADJ ajst-31990 146 72 validation validation NOUN ajst-31990 146 73 model model NOUN ajst-31990 146 74 . . PUNCT ajst-31990 147 1 the the DET ajst-31990 147 2 current current ADJ ajst-31990 147 3 study study NOUN ajst-31990 147 4 is be AUX ajst-31990 147 5 limited limit VERB ajst-31990 147 6 by by ADP ajst-31990 147 7 small small ADJ ajst-31990 147 8 samples sample NOUN ajst-31990 147 9 with with ADP ajst-31990 147 10 a a DET ajst-31990 147 11 single single ADJ ajst-31990 147 12 speech speech NOUN ajst-31990 147 13 modality modality NOUN ajst-31990 147 14 , , PUNCT ajst-31990 147 15 and and CCONJ ajst-31990 147 16 in in ADP ajst-31990 147 17 the the DET ajst-31990 147 18 future future NOUN ajst-31990 147 19 , , PUNCT ajst-31990 147 20 we we PRON ajst-31990 147 21 can can AUX ajst-31990 147 22 integrate integrate VERB ajst-31990 147 23 multimodal multimodal ADJ ajst-31990 147 24 data datum NOUN ajst-31990 147 25 such such ADJ ajst-31990 147 26 as as ADP ajst-31990 147 27 gait gait NOUN ajst-31990 147 28 and and CCONJ ajst-31990 147 29 eeg eeg NOUN ajst-31990 147 30 , , PUNCT ajst-31990 147 31 and and CCONJ ajst-31990 147 32 explore explore VERB ajst-31990 147 33 the the DET ajst-31990 147 34 fusion fusion NOUN ajst-31990 147 35 strategy strategy NOUN ajst-31990 147 36 of of ADP ajst-31990 147 37 logistic logistic ADJ ajst-31990 147 38 regression regression NOUN ajst-31990 147 39 primary primary ADJ ajst-31990 147 40 screening screening NOUN ajst-31990 147 41 and and CCONJ ajst-31990 147 42 xgboost xgboost ADV ajst-31990 147 43 in in ADP ajst-31990 147 44 - - PUNCT ajst-31990 147 45 depth depth NOUN ajst-31990 147 46 analysis analysis NOUN ajst-31990 147 47 . . PUNCT ajst-31990 148 1 through through ADP ajst-31990 148 2 model model NOUN ajst-31990 148 3 collaboration collaboration NOUN ajst-31990 148 4 and and CCONJ ajst-31990 148 5 cross cross ADJ ajst-31990 148 6 - - ADJ ajst-31990 148 7 modal modal ADJ ajst-31990 148 8 learning learning NOUN ajst-31990 148 9 , , PUNCT ajst-31990 148 10 medical medical ADJ ajst-31990 148 11 ai ai NOUN ajst-31990 148 12 will will AUX ajst-31990 148 13 evolve evolve VERB ajst-31990 148 14 from from ADP ajst-31990 148 15 assisted assist VERB ajst-31990 148 16 screening screen VERB ajst-31990 148 17 to to ADP ajst-31990 148 18 intelligent intelligent ADJ ajst-31990 148 19 decision decision NOUN ajst-31990 148 20 - - PUNCT ajst-31990 148 21 making making NOUN ajst-31990 148 22 , , PUNCT ajst-31990 148 23 provide provide VERB ajst-31990 148 24 accurate accurate ADJ ajst-31990 148 25 and and CCONJ ajst-31990 148 26 inclusive inclusive ADJ ajst-31990 148 27 technical technical ADJ ajst-31990 148 28 support support NOUN ajst-31990 148 29 for for ADP ajst-31990 148 30 early early ADJ ajst-31990 148 31 intervention intervention NOUN ajst-31990 148 32 of of ADP ajst-31990 148 33 chronic chronic ADJ ajst-31990 148 34 diseases disease NOUN ajst-31990 148 35 such such ADJ ajst-31990 148 36 as as ADP ajst-31990 148 37 pd pd PROPN ajst-31990 148 38 , , PUNCT ajst-31990 148 39 and and CCONJ ajst-31990 148 40 promote promote VERB ajst-31990 148 41 the the DET ajst-31990 148 42 sinking sinking NOUN ajst-31990 148 43 of of ADP ajst-31990 148 44 medical medical ADJ ajst-31990 148 45 resources resource NOUN ajst-31990 148 46 to to ADP ajst-31990 148 47 the the DET ajst-31990 148 48 grassroots grassroot NOUN ajst-31990 148 49 and and CCONJ ajst-31990 148 50 the the DET ajst-31990 148 51 improvement improvement NOUN ajst-31990 148 52 of of ADP ajst-31990 148 53 diagnostic diagnostic ADJ ajst-31990 148 54 efficiency efficiency NOUN ajst-31990 148 55 . . PUNCT ajst-31990 149 1 references reference NOUN ajst-31990 149 2 [ [ X ajst-31990 149 3 1 1 X ajst-31990 149 4 ] ] PUNCT ajst-31990 149 5 o. o. PROPN ajst-31990 149 6 p. p. PROPN ajst-31990 149 7 neto neto PROPN ajst-31990 149 8 , , PUNCT ajst-31990 149 9 harnessing harness VERB ajst-31990 149 10 voice voice NOUN ajst-31990 149 11 analysis analysis NOUN ajst-31990 149 12 and and CCONJ ajst-31990 149 13 machine machine NOUN ajst-31990 149 14 learning learn VERB ajst-31990 149 15 for for ADP ajst-31990 149 16 early early ADJ ajst-31990 149 17 diagnosis diagnosis NOUN ajst-31990 149 18 of of ADP ajst-31990 149 19 parkinson parkinson NOUN ajst-31990 149 20 's 's PART ajst-31990 149 21 disease disease NOUN ajst-31990 149 22 : : PUNCT ajst-31990 149 23 a a DET ajst-31990 149 24 comparative comparative ADJ ajst-31990 149 25 study study NOUN ajst-31990 149 26 across across ADP ajst-31990 149 27 three three NUM ajst-31990 149 28 datasets dataset NOUN ajst-31990 149 29 , , PUNCT ajst-31990 149 30 journal journal NOUN ajst-31990 149 31 of of ADP ajst-31990 149 32 voice voice PROPN ajst-31990 149 33 , , PUNCT ajst-31990 149 34 journal journal NOUN ajst-31990 149 35 of of ADP ajst-31990 149 36 voice voice NOUN ajst-31990 149 37 , , PUNCT ajst-31990 149 38 2024 2024 NUM ajst-31990 149 39 , , PUNCT ajst-31990 149 40 issn issn PROPN ajst-31990 149 41 0892 0892 NUM ajst-31990 149 42 - - SYM ajst-31990 149 43 1997 1997 NUM ajst-31990 149 44 . . PUNCT ajst-31990 150 1 [ [ X ajst-31990 150 2 2 2 X ajst-31990 150 3 ] ] X ajst-31990 150 4 g. g. PROPN ajst-31990 150 5 solana solana PROPN ajst-31990 150 6 - - PUNCT ajst-31990 150 7 lavalle lavalle PROPN ajst-31990 150 8 , , PUNCT ajst-31990 150 9 r. r. PROPN ajst-31990 150 10 rosas rosas PROPN ajst-31990 150 11 - - PUNCT ajst-31990 150 12 romero romero PROPN ajst-31990 150 13 , , PUNCT ajst-31990 150 14 analysis analysis NOUN ajst-31990 150 15 of of ADP ajst-31990 150 16 voice voice NOUN ajst-31990 150 17 as as ADP ajst-31990 150 18 an an DET ajst-31990 150 19 assisting assisting NOUN ajst-31990 150 20 tool tool NOUN ajst-31990 150 21 for for ADP ajst-31990 150 22 detection detection NOUN ajst-31990 150 23 of of ADP ajst-31990 150 24 parkinson parkinson NOUN ajst-31990 150 25 's 's PART ajst-31990 150 26 disease disease NOUN ajst-31990 150 27 and and CCONJ ajst-31990 150 28 its its PRON ajst-31990 150 29 subsequent subsequent ADJ ajst-31990 150 30 clinical clinical ADJ ajst-31990 150 31 interpretation interpretation NOUN ajst-31990 150 32 , , PUNCT ajst-31990 150 33 biomedical biomedical ADJ ajst-31990 150 34 signal signal NOUN ajst-31990 150 35 processing processing NOUN ajst-31990 150 36 and and CCONJ ajst-31990 150 37 control control NOUN ajst-31990 150 38 , , PUNCT ajst-31990 150 39 2021 2021 NUM ajst-31990 150 40 , , PUNCT ajst-31990 150 41 issn issn PROPN ajst-31990 150 42 1746 1746 NUM ajst-31990 150 43 - - SYM ajst-31990 150 44 8094 8094 NUM ajst-31990 150 45 . . PUNCT ajst-31990 151 1 [ [ X ajst-31990 151 2 3 3 X ajst-31990 151 3 ] ] X ajst-31990 151 4 n. n. PROPN ajst-31990 151 5 d. d. PROPN ajst-31990 151 6 pah pah PROPN ajst-31990 151 7 , , PUNCT ajst-31990 151 8 v. v. PROPN ajst-31990 151 9 indrawati indrawati PROPN ajst-31990 151 10 and and CCONJ ajst-31990 151 11 d. d. PROPN ajst-31990 151 12 k. k. PROPN ajst-31990 151 13 kumar kumar PROPN ajst-31990 151 14 , , PUNCT ajst-31990 151 15 voice voice NOUN ajst-31990 151 16 - - PUNCT ajst-31990 151 17 based base VERB ajst-31990 151 18 svm svm ADJ ajst-31990 151 19 model model NOUN ajst-31990 151 20 reliability reliability NOUN ajst-31990 151 21 for for ADP ajst-31990 151 22 identifying identify VERB ajst-31990 151 23 parkinson parkinson NOUN ajst-31990 151 24 ’s ’s PART ajst-31990 151 25 disease disease NOUN ajst-31990 151 26 , , PUNCT ajst-31990 151 27 ieee ieee NOUN ajst-31990 151 28 access access NOUN ajst-31990 151 29 , , PUNCT ajst-31990 151 30 2023 2023 NUM ajst-31990 151 31 , , PUNCT ajst-31990 151 32 issn issn PROPN ajst-31990 151 33 2169 2169 NUM ajst-31990 151 34 - - PUNCT ajst-31990 151 35 3536 3536 NUM ajst-31990 151 36 . . PUNCT ajst-31990 152 1 [ [ X ajst-31990 152 2 4 4 NUM ajst-31990 152 3 ] ] X ajst-31990 152 4 c.o c.o PROPN ajst-31990 152 5 . . PROPN ajst-31990 152 6 sakar sakar PROPN ajst-31990 152 7 , , PUNCT ajst-31990 152 8 g. g. PROPN ajst-31990 152 9 serbes serbes PROPN ajst-31990 152 10 , , PUNCT ajst-31990 152 11 a. a. NOUN ajst-31990 152 12 gunduz gunduz NOUN ajst-31990 152 13 , , PUNCT ajst-31990 152 14 h.c h.c PROPN ajst-31990 152 15 . . NOUN ajst-31990 152 16 tunc tunc PROPN ajst-31990 152 17 , , PUNCT ajst-31990 152 18 h. h. PROPN ajst-31990 152 19 nizam nizam PROPN ajst-31990 152 20 , , PUNCT ajst-31990 152 21 b.e b.e PROPN ajst-31990 152 22 . . PROPN ajst-31990 152 23 sakar sakar PROPN ajst-31990 152 24 , , PUNCT ajst-31990 152 25 et et PROPN ajst-31990 152 26 al al PROPN ajst-31990 152 27 . . PROPN ajst-31990 152 28 , , PUNCT ajst-31990 152 29 a a DET ajst-31990 152 30 comparative comparative ADJ ajst-31990 152 31 analysis analysis NOUN ajst-31990 152 32 of of ADP ajst-31990 152 33 speech speech NOUN ajst-31990 152 34 signal signal NOUN ajst-31990 152 35 processing processing NOUN ajst-31990 152 36 algorithms algorithm NOUN ajst-31990 152 37 for for ADP ajst-31990 152 38 parkinson parkinson NOUN ajst-31990 152 39 ’s ’s PART ajst-31990 152 40 disease disease NOUN ajst-31990 152 41 classification classification NOUN ajst-31990 152 42 and and CCONJ ajst-31990 152 43 the the DET ajst-31990 152 44 use use NOUN ajst-31990 152 45 of of ADP ajst-31990 152 46 the the DET ajst-31990 152 47 tunable tunable ADJ ajst-31990 152 48 qfactor qfactor NOUN ajst-31990 152 49 wavelet wavelet NOUN ajst-31990 152 50 transform transform NOUN ajst-31990 152 51 , , PUNCT ajst-31990 152 52 applied apply VERB ajst-31990 152 53 soft soft ADJ ajst-31990 152 54 computing computing NOUN ajst-31990 152 55 , , PUNCT ajst-31990 152 56 applied apply VERB ajst-31990 152 57 soft soft ADJ ajst-31990 152 58 computing computing NOUN ajst-31990 152 59 , , PUNCT ajst-31990 152 60 2019 2019 NUM ajst-31990 152 61 , , PUNCT ajst-31990 152 62 issn issn PROPN ajst-31990 152 63 1568 1568 NUM ajst-31990 152 64 - - SYM ajst-31990 152 65 4946 4946 NUM ajst-31990 152 66 . . PUNCT ajst-31990 153 1 [ [ X ajst-31990 153 2 5 5 X ajst-31990 153 3 ] ] PUNCT ajst-31990 153 4 s. s. PROPN ajst-31990 153 5 lahmiri lahmiri PROPN ajst-31990 153 6 , , PUNCT ajst-31990 153 7 a. a. PROPN ajst-31990 153 8 shmuel shmuel PROPN ajst-31990 153 9 , , PUNCT ajst-31990 153 10 detection detection NOUN ajst-31990 153 11 of of ADP ajst-31990 153 12 parkinson parkinson NOUN ajst-31990 153 13 's 's PART ajst-31990 153 14 disease disease NOUN ajst-31990 153 15 based base VERB ajst-31990 153 16 on on ADP ajst-31990 153 17 voice voice NOUN ajst-31990 153 18 patterns pattern NOUN ajst-31990 153 19 ranking rank VERB ajst-31990 153 20 and and CCONJ ajst-31990 153 21 optimized optimize VERB ajst-31990 153 22 support support NOUN ajst-31990 153 23 vector vector NOUN ajst-31990 153 24 machine machine NOUN ajst-31990 153 25 , , PUNCT ajst-31990 153 26 biomedical biomedical ADJ ajst-31990 153 27 signal signal NOUN ajst-31990 153 28 processing processing NOUN ajst-31990 153 29 and and CCONJ ajst-31990 153 30 control control NOUN ajst-31990 153 31 , , PUNCT ajst-31990 153 32 2019 2019 NUM ajst-31990 153 33 , , PUNCT ajst-31990 153 34 issn issn PROPN ajst-31990 153 35 1746 1746 NUM ajst-31990 153 36 - - SYM ajst-31990 153 37 8094 8094 NUM ajst-31990 153 38 . . PUNCT ajst-31990 154 1 [ [ X ajst-31990 154 2 6 6 NUM ajst-31990 154 3 ] ] X ajst-31990 154 4 y.c y.c PROPN ajst-31990 154 5 . . PROPN ajst-31990 154 6 tai tai PROPN ajst-31990 154 7 , , PUNCT ajst-31990 154 8 p.g p.g PROPN ajst-31990 154 9 . . PROPN ajst-31990 154 10 bryan bryan PROPN ajst-31990 154 11 , , PUNCT ajst-31990 154 12 f. f. PROPN ajst-31990 154 13 loayza loayza PROPN ajst-31990 154 14 , , PUNCT ajst-31990 154 15 e. e. PROPN ajst-31990 154 16 pelΓ‘ez pelΓ‘ez PROPN ajst-31990 154 17 , , PUNCT ajst-31990 154 18 a a DET ajst-31990 154 19 voice voice NOUN ajst-31990 154 20 analysis analysis NOUN ajst-31990 154 21 approach approach NOUN ajst-31990 154 22 for for ADP ajst-31990 154 23 recognizing recognize VERB ajst-31990 154 24 parkinson parkinson NOUN ajst-31990 154 25 ’s ’s PART ajst-31990 154 26 disease disease NOUN ajst-31990 154 27 patterns pattern NOUN ajst-31990 154 28 , , PUNCT ajst-31990 154 29 ifac ifac NOUN ajst-31990 154 30 - - PUNCT ajst-31990 154 31 papersonline papersonline NOUN ajst-31990 154 32 , , PUNCT ajst-31990 154 33 2021 2021 NUM ajst-31990 154 34 , , PUNCT ajst-31990 154 35 issn issn PROPN ajst-31990 154 36 2405 2405 NUM ajst-31990 154 37 - - PUNCT ajst-31990 154 38 8963 8963 NUM ajst-31990 154 39 . . PUNCT ajst-31990 155 1 [ [ X ajst-31990 155 2 7 7 X ajst-31990 155 3 ] ] PUNCT ajst-31990 155 4 t. t. PROPN ajst-31990 155 5 j. j. PROPN ajst-31990 155 6 wroge wroge PROPN ajst-31990 155 7 , , PUNCT ajst-31990 155 8 y. y. PROPN ajst-31990 155 9 ΓΆzkanca ΓΆzkanca PROPN ajst-31990 155 10 , , PUNCT ajst-31990 155 11 c. c. PROPN ajst-31990 155 12 demiroglu demiroglu PROPN ajst-31990 155 13 , , PUNCT ajst-31990 155 14 d. d. PROPN ajst-31990 155 15 si si PROPN ajst-31990 155 16 , , PUNCT ajst-31990 155 17 d. d. PROPN ajst-31990 155 18 c. c. PROPN ajst-31990 155 19 atkins atkins PROPN ajst-31990 155 20 and and CCONJ ajst-31990 155 21 r. r. PROPN ajst-31990 155 22 h. h. PROPN ajst-31990 155 23 ghomi ghomi PROPN ajst-31990 155 24 , , PUNCT ajst-31990 155 25 parkinson parkinson NOUN ajst-31990 155 26 ’s ’s PART ajst-31990 155 27 disease disease NOUN ajst-31990 155 28 diagnosis diagnosis NOUN ajst-31990 155 29 using use VERB ajst-31990 155 30 machine machine NOUN ajst-31990 155 31 learning learning NOUN ajst-31990 155 32 and and CCONJ ajst-31990 155 33 voice voice NOUN ajst-31990 155 34 , , PUNCT ajst-31990 155 35 2018 2018 NUM ajst-31990 155 36 ieee ieee NOUN ajst-31990 155 37 signal signal NOUN ajst-31990 155 38 processing processing NOUN ajst-31990 155 39 in in ADP ajst-31990 155 40 medicine medicine NOUN ajst-31990 155 41 and and CCONJ ajst-31990 155 42 biology biology NOUN ajst-31990 155 43 symposium symposium NOUN ajst-31990 155 44 ( ( PUNCT ajst-31990 155 45 spmb spmb PROPN ajst-31990 155 46 ) ) PUNCT ajst-31990 155 47 , , PUNCT ajst-31990 155 48 2018 2018 NUM ajst-31990 155 49 , , PUNCT ajst-31990 155 50 issn issn PROPN ajst-31990 155 51 2473 2473 NUM ajst-31990 155 52 - - PUNCT ajst-31990 155 53 716x 716x NOUN ajst-31990 155 54 . . PUNCT ajst-31990 156 1 [ [ X ajst-31990 156 2 8 8 NUM ajst-31990 156 3 ] ] PUNCT ajst-31990 156 4 p. p. NOUN ajst-31990 156 5 schober schober PROPN ajst-31990 156 6 , , PUNCT ajst-31990 156 7 t.r t.r PROPN ajst-31990 156 8 . . PROPN ajst-31990 156 9 vetter vetter PROPN ajst-31990 156 10 , , PUNCT ajst-31990 156 11 logistic logistic ADJ ajst-31990 156 12 regression regression NOUN ajst-31990 156 13 in in ADP ajst-31990 156 14 medical medical ADJ ajst-31990 156 15 research research NOUN ajst-31990 156 16 . . PUNCT ajst-31990 157 1 anesthesia anesthesia PROPN ajst-31990 157 2 & & CCONJ ajst-31990 157 3 analgesia analgesia PROPN ajst-31990 157 4 132(2 132(2 NUM ajst-31990 157 5 ) ) PUNCT ajst-31990 157 6 , , PUNCT ajst-31990 157 7 2021 2021 NUM ajst-31990 157 8 , , PUNCT ajst-31990 157 9 issn issn PROPN ajst-31990 157 10 0003 0003 NUM ajst-31990 157 11 - - SYM ajst-31990 157 12 2999 2999 NUM ajst-31990 157 13 . . PUNCT ajst-31990 158 1 [ [ X ajst-31990 158 2 9 9 NUM ajst-31990 158 3 ] ] PUNCT ajst-31990 158 4 r. r. PROPN ajst-31990 158 5 santhanam santhanam PROPN ajst-31990 158 6 , , PUNCT ajst-31990 158 7 n. n. PROPN ajst-31990 158 8 uzir uzir PROPN ajst-31990 158 9 , , PUNCT ajst-31990 158 10 s. s. PROPN ajst-31990 158 11 raman raman PROPN ajst-31990 158 12 , , PUNCT ajst-31990 158 13 s. s. PROPN ajst-31990 158 14 banerjee banerjee PROPN ajst-31990 158 15 , , PUNCT ajst-31990 158 16 experimenting experiment VERB ajst-31990 158 17 xgboost xgboost ADV ajst-31990 158 18 algorithm algorithm NOUN ajst-31990 158 19 for for ADP ajst-31990 158 20 prediction prediction NOUN ajst-31990 158 21 and and CCONJ ajst-31990 158 22 classification classification NOUN ajst-31990 158 23 of of ADP ajst-31990 158 24 different different ADJ ajst-31990 158 25 datasets dataset NOUN ajst-31990 158 26 , , PUNCT ajst-31990 158 27 2017 2017 NUM ajst-31990 158 28 , , PUNCT ajst-31990 158 29 issn issn PROPN ajst-31990 158 30 0974–5572 0974–5572 NUM ajst-31990 158 31 . . PUNCT ajst-31990 159 1 [ [ X ajst-31990 159 2 10 10 NUM ajst-31990 159 3 ] ] X ajst-31990 159 4 d.m d.m PROPN ajst-31990 159 5 . . PUNCT ajst-31990 159 6 abdullah abdullah PROPN ajst-31990 159 7 , , PUNCT ajst-31990 159 8 a.m. a.m. PROPN ajst-31990 159 9 abdulazeez abdulazeez PROPN ajst-31990 159 10 , , PUNCT ajst-31990 159 11 machine machine NOUN ajst-31990 159 12 learning learning NOUN ajst-31990 159 13 applications application NOUN ajst-31990 159 14 based base VERB ajst-31990 159 15 on on ADP ajst-31990 159 16 svm svm ADJ ajst-31990 159 17 classification classification NOUN ajst-31990 159 18 a a DET ajst-31990 159 19 review review NOUN ajst-31990 159 20 . . PUNCT ajst-31990 160 1 qubahan qubahan PROPN ajst-31990 160 2 academic academic ADJ ajst-31990 160 3 journal journal NOUN ajst-31990 160 4 , , PUNCT ajst-31990 160 5 2021 2021 NUM ajst-31990 160 6 , , PUNCT ajst-31990 160 7 issn issn PROPN ajst-31990 160 8 2709 2709 NUM ajst-31990 160 9 - - SYM ajst-31990 160 10 8206 8206 NUM