Title Science and Technology Indonesia e-ISSN:2580-4391 p-ISSN:2580-4405 Vol. 7, No. 3, July 2022 Research Paper Pattern Recognition using Multiclass Support Vector Machine Method with Local Binary Pattern as Feature Extraction Nursyiva Irsalinda1*, Sugiyarto Surono1*, Indah Dwi Ratna Sary1 1Department of Mathematics, Ahmad Dahlan University, Yogyakarta, 55191, Indonesia *Corresponding author: nursyiva.irsalinda@math.uad.ac.id, sugiyarto@math.uad.ac.id AbstractPattern recognition is a scientific discipline usually used to classify objects into a number of categories or classes through a featureextraction method applied to recognize an object accurately. Meanwhile, Local Binary Pattern (LBP) is a texture analysis methodwhich uses statistical and structural models for feature extraction. Moreover, a Support Vector Machine (SVM) method is normallyused to solve non-linear problems in high dimensions to obtain an optimal solution by finding the best hyperplane through themaximizing of the margin between two data classes. Pattern recognition in paintings using machine learning has never been done inany research. Meanwhile it is very important in the future to be able to serve as a verification system for novelty works of art at thestage of filing for intellectual property rights. Therefore, this study aimed to apply pattern recognition with LBP feature extractionmethod and multiclass SVM classification method to classify the flow of several classes of painting works including expressionism,fauvism, naturalism, realism, and romanticism. The best evaluation results using this method were obtained in the training andtesting data combination of 90:10 with an accuracy rate of 83%. Therefore, it can be concluded that machine learning in patternrecognition of painting works can be applied. KeywordsPattern Recognition, Local Binary Pattern, Support Vector Machine, Multiclass SVM Received: 19 January 2022, Accepted: 23 April 2022 https://doi.org/10.26554/sti.2022.7.3.269-274 1. INTRODUCTION Technological developments have made data become one of the important components in every human activity. These data can be numeric, categorical, text, images, and others but the most widely used form in recent times is the images which are usually processed to produce a good identication pro- cess. Meanwhile, an important method usually used in solving problems in image processing is pattern recognition which is a scientic discipline designed to classify objects into a number of categories or classes (Theodoridis, 2009). Moreover, the numerical pattern recognition which is a statistical approach in- volves using multidimensional data vector as input where each component is called a feature. The numerical pattern recog- nition observed to be increasingly used in decision-making processes. Furthermore, a pattern recognition process through feature extraction method such as the Local Binary Pattern method is needed to recognize an object accurately. Local Binary Pattern (LBP) is a texture analysis method which uses statistical and structural models for feature extrac- tion (Amat et al., 2017; Nugraha et al., 2019). It works by applying a threshold to the surrounding pixels based on the pixels being observed (Galshetwar et al., 2017; Sharma et al., 2020) and also reported to be widely used in applications in- volving image texture capture and classication (Kaewchote et al., 2018; Prashanth et al., 2020). Some of its advantages includes being xed to changes in light intensity from the same object and accuracy in recognizing an object (Susanto et al., 2018). Moreover, it can also be implemented easily and fairly quickly for feature extraction using low computational pro- cesses (Amat et al., 2017). Classication is one of the supervised learning methods in data mining. It is usually used to nd a model or function of a data set by grouping it into classes such that the data objects determined in each class label indicates they have the same characteristics or similar pattern (Cervantes et al., 2020). The grouping process involves dividing the data into predetermined classes and assigning the classes according to the similarity in characteristics and the patterns in the words (Styawati and Mustofa, 2019). Support Vector Machine (SVM) was introduced by Vapnik in 1995 is a method widely used in classifying supervised learn- ing (Ay et al., 2019; Komarudin et al., 2020; Wang and Zhao, https://crossmark.crossref.org/dialog/?doi=10.26554/sti.2022.7.3.269-274&domain=pdf https://doi.org/10.26554/sti.2022.7.3.269-274 Irsalinda et. al. Science and Technology Indonesia, 7 (2022) 269-274 2020; Yin and Yin, 2016). SVM is formulated to solve non- linear problems in high dimensions to obtain optimal solutions (Li et al., 2020; Liu et al., 2017; Pramudita and Musdholifah, 2020; Styawati and Mustofa, 2019; Tang et al., 2019). This methodnds thebesthyperplanebymaximizingthemarginbe- tween two classes of data (Komarudin et al., 2020). Meanwhile, the margin is dened as the distance between the hyperplane and the support vector which is the closest data from the class to the hyperplane. It is important to note that SVM uses a trick Kernel function to calculate data with non-linear prob- lems in order to obtain classication results with a good level of accuracy. However, it is sensitive to noise and outliers and also reported is less ecient in obtaining complex models for large-scale data sets (Cervantes et al., 2020). Meanwhile, some of its benets include the ability to minimize errors and higher accuracy when compared with other classications methods (Utami et al., 2021). Pattern recognition and SVM classication methods have been widely used in previous studies. For example, Dias Aziz Pramudita and Aina Musdholifah added a parameter optimiza- tion process known as the Gravitational Search Algorithm (GSA) to improve the accuracy of the SVM method in the thyroid nodule classication process (Pramudita and Musdho- lifah, 2020). L. Jerlin Rubini and Eswaran Perumal also used the Fruit Fly Optimization Algorithm (FFOA) to select the best features in medical data which were later processed using the multiclass SVM method for classication (Jerlin Rubini and Perumal, 2020). Moreover, Diya Wang and Yixi Zhao predicted the feelings of investors towards stock market news using the SVM method in order to make decisions with a high level of accuracy (Wang and Zhao, 2020). Study about SVM also developed to enhance age invariant face recognition perfor- mance based on gender classication (Nayak and Indiramma, 2021) while Mayur succeeded in showing that CNN-ECOC produced higher accuracy than ordinary CNN classication (Bora et al., 2020). Another research found a higher accuracy of performance for multiclass SVM method when compared with SVM method (Alita et al., 2020). from the studies related to pattern recognition that have been described. There has been no research on the multiclass SVM method that uses LBP as its feature extraction. By looking at the advan- tages of LBP and in order to nd a method with better results, the researchers combined the multiclass SVM method with LBP to classify pattern recognition. The case study adopted in this research is the classication of paintings into ve classes, namely expressionism, fauvism, naturalism, realism, and ro- manticism. This case study was appointed because pattern recognition in paintings using machine learning has never been done in any research. Meanwhile it is very important in the future to be able to serve as a verication system for novelty works of art at the stage of ling for intellectual property rights. This present study, therefore, uses the Multiclass Support Vec- tor Machine (MSVM) method to classify pattern recognition using the LBP method for feature extraction. The aim was to determine the eectiveness of applying these methods in classifying the ow in a painting. 2. MATERIALS AND METHODS 2.1 Pattern Recognition Patternrecognition isadisciplineusuallyusedtoclassifyobjects intoanumberofcategoriesorclasses (Theodoridis,2009). The object can be an image, a signal waveform, or any type of mea- surement that needs to be classied. This means pattern recog- nition is an important part of a system for decision-making purposes. 2.2 Local Binary Pattern Method The Local BinaryPattern method was rst introduced byTimo Ojala in 1994 as a texture analysis method which uses statistical and structural models (Ojala et al., 2000). It also has the ability to compare the gray values of the surrounding pixels (Abbas et al., 2019; Agbele et al., 2019; Ali et al., 2017; Kwabena et al., 2020; Muqeet and Holambe, 2019; Vidya and Chandra, 2019). The basic LBP operator was congured using 8 pixels around a center pixel and the threshold of the nth surrounding pixel was through the gray value of the center pixel and the thresholding function. The binary code generated by this LBP was, therefore, used to represent the features of the center pixel ic. LBP(xc, yc) = n−1∑︁ n=0 s(in − ic)2n (1) s(x) = { 1, x ≥ 0 0, x < 0 (2) where xc = pixel width yc = pixel height s(x) = threshold function which is dened by (2) in = ambient pixels from the center ic = center pixel LBP is easy to implement, has lower computational level, does not require a long time for feature extraction (Shiyam, 2019), xed to changes in light intensity from the same object, and accurate in recognizing an object (Gonzalez, 2009). 2.3 Support Vector Machine Method Support Vector Machine (SVM) is one of the methods usually used classifying supervised learning and reported to be rst introduced by Vapnik in 1995 (Ay et al., 2019; Komarudin et al., 2020; Wang and Zhao, 2020; Yin and Yin, 2016) to solve non-linear problems at high dimensions (Abbas et al., 2019; Battineni et al., 2019) in order to obtain an optimal solution. The method nds the best hyperplane by maximizing the margin between two data classes. Meanwhile, a hyperplane is dened as a line or eld separating the data between dierent classes while margin is the distance between the hyperplane and the support vector which acts as the closest data from the class to the hyperplane. © 2022 The Authors. Page 270 of 274 Irsalinda et. al. Science and Technology Indonesia, 7 (2022) 269-274 For example, the training data set D is Dn = (x1, y1), (x2, y2), ..., (xn, yn), where D= b{(xi , yi)}ni=1, xi ∈ R d is the training data set while x1 is the input data and y1 ∈ (+1,-1) is a class label with an output value of +1 (positive) or -1 (negative). The main concept used in assigning the separator to a lin- ear separable is the dot product between two vectors which is dened as wTx= Σiwixi. Where, w is the weighting vector and b is the bias. Moreover, the separating hyperplane was used to divide the space into two classes as indicated in the following relationship. h(x) = wTxi + b + 0 (3) Meanwhile, the separator function for the two classes is as follows wTxi + b ≥ +1 for yi = +1 and wTxi + b ≤ −1 for yi = −1 where, |b|| |w| | is the perpendicular distance of the dividing plane from the center of the coordinates and ||w|| is the Euclidean distance (Euclidean norm) from w. Moreover, the Quadratic Programming (QP) problem was thestepusedindeterminingtheminimumpointbyconsidering the equation constraints. min w,b 1 2 ||w||2 (4) with constraint function yi (wxi + b) − 1 ≥ 0, i = 1, 2, ..., m Itisalsopossible toresolvetheoptimizationproblemthrough the use of the Lagrange multiplier as indicated in the following equation: L = 1 2 ||w||2 − n∑︁ i=1 𝛼i (yi ((wxi + b) − 1)) (5) Where, 𝛼i is the Lagrange Multiplier which is zero or positive (𝛼i ≥ 0). The problem was, therefore, optimized by minimiz- ing L to w and b and maximizing 𝛼i. The equation was later modied as follows with due consideration for the nature of the optimal point L= 0: Condition 1 𝜕Lp 𝜕w = 0 → w = n∑︁ i=1 𝛼iyixi (6) Condition 1 𝜕Lp 𝜕b = 0 → w = n∑︁ i=1 𝛼iyi = 0 (7) The substitution of w changed the equation into a Lagrange duality Max LD = n∑︁ i=1 𝛼i − 1 2 ∑︁ i, j 𝛼i𝛼 jyiyjxi · xj (8) Condition 1 n∑︁ i=1 𝛼iyi = 0 Condition 2 𝛼i ≥ 0, i = 1, 2, ..., n 2.4 Kernel The Support Vector Machine learning process to determine the support vectors depends on the dot product of the data in the feature space i.e. Φi(xi) ·Φi(xj). It is, therefore, possible to replace this dot product calculations with trick Kernel func- tions K(xi ,xj) which implicitly dene the transformation and formulated as follows: K(xi , xj) = Φi (xi) · Φ(xj) (9) One of the trick Kernel functions commonly used in the general classication of Support Vector Machines is Radial Basis Function (RBF) which is presented as follows. K(xi , xj) = exp ( − ||xi − xj ||2 2𝜎2 ) (10) Previous studies have alreadyshown that the Radial Basis Func- tion (RBF) Kernel has a better accuracy rate than other Kernels (Caraka et al., 2017). Therefore, it was applied in this research for classication using the Multiclass Support Vector Machine method. 2.5 Multiclass SVM The Support Vector Machine (SVM) method was rst intro- duced by Vapink only to classify data into two classes. Subse- quent research has, however, been conducted to increase its ability to classify data for more than two classes. Moreover, there are two approaches to solve SVM problems for multiclass and the rst includes the combination of all data in an optimiza- tion problem while the second involves building a multiclass classier or several binary SVMs. The rst approach requires solving a more complex and computationally high optimiza- tion problem and this limits its development. Meanwhile, the second such as "One-Against-All" approach is usually used to implement multiclass SVM. This One-Against-All (OAA) method is usually used to determine the k binary SVM models where k is the number of classes (Chamasemani and Singh, 2011; Kumar and Gopal, 2011). It also tests p l with all data from the class with label +1 and those from other classes with label -1. An example of a classication problem with ve classes and ve binary SVMs is presented on Table 1. © 2022 The Authors. Page 271 of 274 Irsalinda et. al. Science and Technology Indonesia, 7 (2022) 269-274 Table 1. OAA Method yi= 1 yi= -1 Hypothesis Class 1 Not class 1 f1(x)= (w1)x+b1 Class 2 Not class 2 f2(x)= (w2)x+b2 Class 3 Not class 3 f3(x)= (w3)x+b3 Class 4 Not class 4 f4(x)= (w4)x+b4 Class 5 Not class 5 f5(x)= (w5)x+b5 3. RESULTS AND DISCUSSION This research used painting works data including expression- ism, fauvism, naturalism, realism, and romanticism. A total of 420 images on ow of painting were obtained and these in- clude 70 for expressionism, 100 for fauvism, 80 for naturalism, 70 for realism, and 100 for romanticism. These image data were later divided into training and testing samples at dierent combinations including 90:10, 75:25, 50:50, and 25:75. Table 2. LBP Value Data Value 0 1 2 3 4 5 . . . 255 Image 1 833 117 99 79 126 49 . . . 937 Image 2 676 86 272 85 86 12 . . . 807 Image 3 754 110 96 76 95 42 . . . 920 ... ... ... ... ... ... ... . . . ... Image 420 393 67 226 110 57 14 . . . 603 Figure 1. Prediction Results and The Actual Data 3.1 Result of Feature Extraction Method LBP Each of the images was resized and converted to a grayscale in the LBP to determine the pixel values at each point. These val- ues were made into threshold and their binary values encoded to produce LBP values for all pixel points in the image. The nal result of the feature extraction is presented on a histogram and one image was discovered to have 256 values based on the frequency of occurrence from 0 to 255. These outputs are also indicated in the on Table 2. The data from the LBP extraction process were used as the input for the classication conducted through the Multiclass Support VectorMachine (MSVM) using the Radial Basis Func- tion (RBF) kernel. The results of the prediction against the actual data for each of the proportions are, therefore, shown in the following graphic images: Figure 1 shows the 90:10 combination produced predic- tions which are almost the same as the actual data. Moreover, the classication results for each training and testing data com- bination were evaluated using a confusion matrix to determine the deviations in the prediction process and the ndings are presented as follows. Figure 2. (a) Confusion Matrix Generated by 90:10 Partitio- ned Data. (b) Confusion Matrix Generated by 75:25 Parti- tioned Data. (c) Confusion Matrix Generated by 50:50 Par- titioned Data. (d) Confusion Matrix Generated by 25:75 Par- titioned Data Figure 2 shows the correct prediction results are located in the diagonal part of the table. Moreover, the performance of the MSVM classication model was determined based on precision, recall, and accuracy as indicated in Figure 3: Figure 3. Evaluation Model on Each Data Partition © 2022 The Authors. Page 272 of 274 Irsalinda et. al. Science and Technology Indonesia, 7 (2022) 269-274 Figure 3 illustrates the evaluation classication that has been produced by the combination of the LBP method with MSVM to classify the data of works of art in the form of paint- ing. From the four partition combination experiments that is 90:10, 75:25, 50:50 and 25:75 the accuracy, precision and recall values are obtained as shown in Figure 3. From the bar charts, it can be seen clearly that the data partition is 90:10 more has the highest value in terms of accuracy, precision and recall. While the 25:75 data partition has the lowest model evaluation value among the others. 4. CONCLUSIONS Data analysis and discussion showed that pattern recognition through Local Binary Pattern feature extraction and Multiclass Support Vector Machine were used to classify works of art into ve classes which include expressionism, fauvism, natu- ralism, realism, and romanticism. It was discovered that the classication made by combining training and testing data at 90:10 had an accuracy rate of 83%, 75:25 had 81%, 50:50 had 81%, and 25:75 had 51%. Therefore, the best evaluation of pattern recognition classication in this data case using Mul- ticlass Support Vector Machine with Local Binary Pattern as feature extraction were obtained through training and testing data combined at 90:10. Based on the results of the study, it was concluded that the pattern recognition process using the Local Binary Pattern method for feature extraction and the Multiclass Support Vector Machine method can be used to classify works of art into ve classes, namely expressionism, fauvism, naturalism, realism, and romanticism. Therefore, this study needs to be further developed by adding more types of paintings to be used as a novelty verication system for works of art at the stage of ling intellectual property rights. 5. ACKNOWLEDGMENT The author thanks to research and community service insti- tute of Ahmad Dahlan University who give nancial support through ocial research scheme funding. REFERENCES Abbas, A. K., N. A. Al-haideri, and A. A. Bashikh (2019). Im- plementing Articial Neural Networks and Support Vector Machines to Predict Lost Circulation. Egyptian Journal of Petroleum, 28(4); 339–347 Agbele, T., B. Ojeme, and R. Jiang (2019). Application of Local Binary Patterns and Cascade AdaBoost Classier for Mice Behavioural Patterns Detection and Analysis. Procedia Computer Science, 159; 1375–1386 Ali, M. M., P. L. Yannawar, and A. Gaikwad (2017). Multi- Algorithm of Palmprint Recognition System based on Fu- sion of Local Binary Pattern and Two-Dimensional Locality Preserving Projection. Procedia Computer Science, 115; 482– 492 Alita, D., Y. Fernando, and H. Sulistiani (2020). Implementasi Algoritma Multiclass SVM pada Opini Publik Berbahasa Indonesia di Twitter. Jurnal Tekno Kompak, 14(2); 86–91 (in Indonesia) Amat, R., J. Y. Sari, and I. P. Ningrum (2017). Implementasi Metode Local Binary Patterns untuk Pengenalan Pola Huruf Hiragana dan Katakana pada Smartphone. JUTI: Jurnal Ilmiah Teknologi Informasi, 15(2); 152 (in Indonesia) Ay, M., D. Stenger, M. Schwenzer, D. Abel, and T. Bergs (2019). Kernel Selection for Support Vector Machines for System Identication of a CNC Machining Center. IFAC- PapersOnLine, 52(29); 192–198 Battineni, G., N. Chintalapudi, and F. Amenta (2019). Ma- chine Learning in Medicine: Performance Calculation of Dementia Prediction by Support Vector Machines (SVM). Informatics in Medicine Unlocked, 16; 100200 Bora, M. B., D. Daimary, K. Amitab, and D. Kandar (2020). Handwritten Character Recognition from Images using CNN-ECOC. Procedia Computer Science, 167; 2403–2409 Caraka, B., B. A. A. Sumbodo, and I. Candradewi (2017). Klasikasi Sel Darah Putih Menggunakan Metode Support Vector Machine (SVM) Berbasis Pengolahan Citra Digital. IJEIS (Indonesian Journal of Electronics and Instrumentation Systems), 7(1); 25–36 (in Indonesia) Cervantes, J., F. Garcia-Lamont, L. Rodríguez-Mazahua, and A. Lopez (2020). A Comprehensive Survey on Support Vector Machine Classication: Applications, Challenges and Trends. Neurocomputing, 408; 189–215 Chamasemani, F. F. and Y. P. Singh (2011). Multi-Class Sup- port Vector Machine (SVM) Classiers-an Application in Hypothyroid Detection and Classication. In 2011 6th In- ternational Conference on Bio-Inspired Computing: Theories and Applications. Institute of Electrical and Electronics Engineers; 351–356 Galshetwar, G. M., L. M. Waghmare, A. B. Gonde, and S. Mu- rala (2017). Edgy Salient Local Binary Patterns in Inter- Plane Relationship for Image Retrieval in Diabetic Retinopa- thy. Procedia Computer Science, 115; 440–447 Gonzalez, R. C. (2009). Digital Image Processing. Pearson Education India Jerlin Rubini, L. and E. Perumal (2020). Ecient Classication of Chronic Kidney Disease by using Multi-Kernel Support Vector Machine and Fruit Fly Optimization Algorithm. In- ternational Journal of Imaging Systems and Technology, 30(3); 660–673 Kaewchote, J., S. Janyong, and W. Limprasert (2018). Image Recognition Method using Local Binary Pattern and The Random Forest Classier to Count Post Larvae Shrimp. Agriculture and Natural Resources, 52(4); 371–376 Komarudin, S., D. Anggraeni, A. Riski, and A. Hadi (2020). Classication of Genetic Expression in Prostate Cancer us- ing Support Vector Machine Method. In Journal of Physics: Conference Series, 1613; 012032 Kumar, M. A. and M. Gopal (2011). Reduced One-Against- All Method for Multiclass SVM Classication. Expert Systems with Applications, 38(11); 14238–14248 Kwabena, P. M., B. A. Weyori, and A. A. Mighty (2020). Ex- © 2022 The Authors. Page 273 of 274 Irsalinda et. al. Science and Technology Indonesia, 7 (2022) 269-274 ploring The Performance of LBP-Capsule Networks with K-Means Routing on Complex Images. Journal of King Saud University-Computer and Information Sciences Li, L. L., X. Zhao, M. L. Tseng, and R. R. Tan (2020). Short- Term Wind Power Forecasting based on Support Vector Machine with Improved Dragony Algorithm. Journal of Cleaner Production, 242; 118447 Liu, C., W. Wang, M. Wang, F. Lv, and M. Konan (2017). An Ecient Instance Selection Algorithm to Reconstruct Training Set for Support Vector Machine. Knowledge-Based Systems, 116; 58–73 Muqeet, M. A. and R. S. Holambe (2019). Local Binary Pat- terns based on Directional Wavelet Transform for Expres- sion and Pose-Invariant Face Recognition. AppliedComputing and Informatics, 15(2); 163–171 Nayak, J. S. and M. Indiramma (2021). An Approach to En- hance Age Invariant Face Recognition Performance based on Gender Classication. Journal of King Saud University- Computer and Information Sciences Nugraha, E., A. Wibawa, M. Hakim, U. Kholifah, R. Dini, and M. Irwanto (2019). Implementation of Fuzzy Tsukamoto Method in Decision Support System of Journal Acceptance. In Journal of Physics: Conference Series, 1280; 022031 Ojala, T., M. Pietikäinen, and T. Mäenpää (2000). Gray Scale and Rotation Invariant Texture Classication with Local Binary Patterns. In European Conference on Computer Vision, 1842; 404–420 Pramudita, D. A. and A. Musdholifah (2020). GSA to Obtain SVM Kernel Parameter for Thyroid Nodule Classication. IJCCS (Indonesian Journal of Computing and Cybernetics Sys- tems), 14(1); 11–22 Prashanth, D. S., R. V. K. Mehta, and N. Sharma (2020). Clas- sicationofHandwrittenDevanagariNumber-anAnalysisof Pattern Recognition Tool using Neural Network and CNN. Procedia Computer Science, 167; 2445–2457 Sharma, A., A. Mittal, S. Singh, and V. Awatramani (2020). Hand Gesture Recognition using Image Processing and Fea- ture Extraction Techniques. Procedia Computer Science, 173; 181–190 Shiyam, A. L. (2019). PresensiMahasiswaDenganEkstraksiFitur Wajah Menggunakan Metode Local Binary Pattern. Universitas Nusantara PGRI Kediri (in Indonesia) Styawati,S.andK.Mustofa(2019). ASupportVectorMachine- Firey Algorithm for Movie Opinion Data Classication. IJCCS (Indonesian Journal of Computing and Cybernetics Sys- tems), 13(3); 219–230 Susanto, A., D. Sinaga, C. A. Sari, E. H. Rachmawanto, and D. Setiadi (2018). A High Performace of Local Binary Pat- tern on Classify Javanese Character Classication. Scientic Journal of Informatics, 5(1); 8 Tang, L., X. Lu, C. Yang, and X. Li (2019). Extension Sup- port Vector Machine for Extension Classication. Procedia Computer Science, 162; 174–182 Theodoridis, S. (2009). Pattern Recognition. Elsevier Utami, N. A., W. Maharani, and I. Atastina (2021). Person- ality Classication of Facebook users According to Big Five Personality using SVM (Support Vector Machine) Method. Procedia Computer Science, 179; 177–184 Vidya, B. S. and E. Chandra (2019). Entropy based Local Binary Pattern (ELBP) Feature Extraction Technique of Multimodal Biometrics as Defence Mechanism for Cloud Storage. Alexandria Engineering Journal, 58(1); 103–114 Wang, D. and Y. Zhao (2020). Using News to Predict Investor Sentiment: Based on SVM Model. Procedia Computer Science, 174; 191–199 Yin, S. and J. Yin (2016). Tuning Kernel Parameters for SVM based on Expected Square Distance Ratio. Information Sci- ences, 370; 92–102 © 2022 The Authors. Page 274 of 274 INTRODUCTION MATERIALS AND METHODS Pattern Recognition Local Binary Pattern Method Support Vector Machine Method Kernel Multiclass SVM RESULTS AND DISCUSSION Result of Feature Extraction Method LBP CONCLUSIONS ACKNOWLEDGMENT