Format Template


 

 

Vol. 3, No. 2 | Jul – Dec 2019 
 

 

 

SJCMS | P-ISSN: 2520-0755| E-ISSN: 2522-3003 © 2019 Sukkur IBA University – All Rights Reserved 

1 

 

Performance Comparison of SVM and its Variants for the 

Early Prognosis of Breast Cancer 

Talha Ahmed Khan1,2, Muhammad Alam3,4, Zeeshan Shahid 5 , M.S. Mazliham6 

 

Abstract: 

Breast cancer has become a leading cause of women death in this era. Breast cancer is very 

common in various countries including Pakistan. Early identification of the breast cancer or tumor 

is the only way for the rapid treatment and cure. An imaging approach named as mammography 

has performed tremendous job in the field of medical to detect the cancer tumors on early basis 

with less false alarm rate. Breast cancer has two types of tumors a) Benign and b) Malignant. 

Malignant is acknowledged as cancer tumor as it spread and grow rapidly inside the tissues. 

Detection of Malignant tumor is very complex in dense breast as it is covered and linked with the 

milk glands, ducts and other related tissues. Therefore, machine learning and artificial intelligence 

approaches were needed as mammographic images required edge detection, image enhancement 

and image processing. Various Artificial Intelligence based algorithms have been applied to the 

clinical breast cancer data set for the early detection of breast tumor. In this research work the 

clinical data has been collected from the UCI machine learning repository for the classification 

of breast cancer tumor a) Benign and b) Malignant. Support Vector machine with its variants 

Kernel, Gaussian Kernel and Sigmoid Kernel have been applied to the linearly separable breast 

cancer data set for comparative analysis. Results proved that all the variants of SVM performed 

better for the breast cancer classification.   

Keywords: Benign, Malignant, Breast cancer, dense breast, fatty breast, Support Vector 

Machine 

1. Introduction  

Digital images based elasto tomography 
(DIET) was proposed for the robust evaluation 
of breast tumor. Vibrated breast image was 
captured and surface motion was recorded to 
track the tumor.  3D surface motion was 

                                                           
1 British Malaysian Institute (BMI), Universiti Kuala Lumpur, Malaysia  
2 Usman Institute of Technology, Karachi, Pakistan 
3 CCIS, Institute of Business Management, Karachi, Pakistan  
4
Malaysian Institute of Information and Technology (MIIT), Universiti Kuala Lumpur Malaysia 

5 Electrical Engineering Department, Institute of Business Management, Karachi, Pakistan  
6 Malaysian France Institute (MFI), Universiti Kuala Lumpur Malaysia 

 

accurately evaluated using this approach. 
DIET based approach could detect upto 100 
mm. of tumor [1]. Gene expression pattern 
data sets were used to find out the similarities 
and differences between breast tumor and 
peripheral blood monocular cells [2]. An 
intelligent system was proposed comprised of 



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

2 

 

neural network classifier and image processing 
(ISIBC). The features for the classification 
have been obtained by applying GLCM 
algorithm. Standard deviation, mean and 
entropy were calculated. Filtering, edge 
detection and morphological techniques were 
applied to the images for the feature’s 
extraction [3]. Mammographic images were 
collected from the data bases available on the 
internet [3]. Image processing method were 
applied to filter the noise and to enhance the 
mammographic images for the robust 
investigation of breast cancer. Adaptive 
histogram equalization, wavelets and data 
fusion methods were applied to identify micro-
calcifications using LabView [4]. 
Reproduction of cells is the main indication of 
the breast cancer. Reproduction can usually be 
examined by the color of the cells for the 
antigen Ki-67. Artificial Neural network 
comprising of the development of multi-layer 
perceptron network with the combination of 
feed forward neural network was proposed for 
the accurate and precise classification of three 
different cancer genes which are linked with 
Ki-67. Results proved that unique non-
recursive method was very successful to 
provide the complete detailed genes network 
with the visualizations [5]. 

 
Fig. 1. Mammogram image of breast tumor 

in fatty and dense breast.  

Fig. 1 demonstrated the downloaded image of 

the mammogram which has been downloaded 

from the mammographic images databases 

available on the internet. The figure explains 

that it is very easy to determine the malignant 

tumor in a fatty breast compared to the dense 

breast as it is very complex to identify the 

malignant tumor in the dense forest. Even a 

big tumor is very difficult to analyze in a dense 

breast. Therefore, image processing and 

filtering techniques are needed to enhance the 

image quality and features. Mammography is 

considered as the widely used tool for the 

detection of the malignant tumor in the breast 

but in dense breast tissues may overlap each 

other which causes the confusion for exact 

investigation of malignant tumor. Tissue 

overlapping may be minimized by using 

digital tomosynthesis. Dense breast can be 

defined as the combination of milk ducts, 

glands and linked tissues. 4-D virtual 

prototype breast were designed that were 

compared to the fourteen imaging approaches 

[6]. Mammographic filtered images were 

divided into small grids and features were 

extracted from these small grids for better 

classification of Malignant and Benign. 

Results proved that grid-based pattern 

evaluation for the doubtful tumor or tissue 

performed better with the efficiency of 

91.67%. The results were also compared with 

the knowledge-based data base provided by 

the radiologist [7]. OTSU method was also 

adopted to reduce the noise and segment the 

mammogram images to discriminate the 

changes compared to the normal one [8]. 

Simplest evolving connectionist system was 

suggested for the classification of normal, 

benign and malignant classification. 

Wisconsin dataset was collected from UCI 

machine learning repository and sensitivity of 

96.02% was achieved [9]. Computer aided 

diagnostic based approach was developed and 

82% sensitivity was achieved [10]. 

2. Problem Statement  

Women normally take the cancer tumor 

symptoms very lightly due to the the lack of 

awareness for the malignant tumor which 

spreads and grow inside the breast tissues. 

Vigorous classification of malignant and 

Benign tumor can be acknowledged as very 

complex due to the similarities between them 



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

3 

 

which creates confusion [11-14]. 

Mammography is the most widely used 

technique for the screening of breast cancer 

but image filtering, processing and machine 

learning techniques were needed for the 

accurate results as identification of malignant 

tumor in dense breast is very difficult. Skin 

inflammation and breast cancer have some 

common symptoms. Breast pain, swelling, 

and change of skin color are very common 

symptoms of cancer but people ignore it as 

they take it as normal skin inflammatory 

problem [15-17].  

 

3. Methodology 

3.1. Fundamental Block Diagram 

 

 
Fig. 2. Fundamental Block diagram 

Fig. 1 demonstrated the fundamental block 

diagram for the breast cancer classification. 

Breast Cancer tumor can be classified into 

Benign and Malignant. Malignant is the breast 

tumor which can grow with the tissues and 

spread rapidly. Breast cancer data acquisition 

was performed and then the data was 

normalized as the data contained some missed 

and repetitive values. SVM classifier with the 

variants of Kernel, Sigmoid and Gaussian 

Kernel were applied to the breast cancer data 

set for the classification of Benign and 

Malignant tumor. The results were compared 

with each other variants of the SVM. SVM 

proved to be the competent classifier. 

 

3.2. Breast cancer Dat collection 

Table I: Breast cancer data set [18-19] 

 
Table no. 1 displayed all the attributes related 

to the identification of the breast tumor. 

Attributes can be explained as: 

1. Sample code number: id number  

2. Clump Thickness: 1 - 10  

3. Uniformity of Cell Size: 1 - 10  

4. Uniformity of Cell Shape: 1 - 10  

5. Marginal Adhesion: 1 - 10  



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

4 

 

6. Single Epithelial Cell Size: 1 - 10  

7. Bare Nuclei: 1 - 10  

8. Bland Chromatin: 1 - 10  

9. Normal Nucleoli: 1 - 10  

10. Mitoses: 1 - 10  

11. Class: (2 for benign, 4 for malignant) 

 
Fig. 3. Imported data set in Python 

 

Data set was imported in python and then 

normalized to filter the data. Data 

normalization was performed as data 

contained some missed and repetitive values. 

4. Algorithm Implementation 

4.1. Support Vector Machine (SVM)  

Support Vector Machine classifier 
algorithm was applied to the breast cancer data 
set for the accurate classification of brain 
tumor. Pandas and Numpy libraries were 
imported in the Python environment for the 
implementation of SVM. SVM algorithm 
creates a boundary by using hyperplanes for 
the classification. Simple SVM is very useful 
to classify the linear data. Two hyperplanes 
were created for this data classification. 

H1, H2 are the hyper planes:  

H1: w•xi +b = 2                                           (1)                                               

H2: w•xi +b = 4                                           (2) 

SVC(C=1.0, cache_size=200, 

class_weight=None, coef0=0.0, 

decision_function_shape='ovr', degree=3, 

gamma='auto_deprecated', 

kernel='linear', max_iter=-1, 

probability=False, random_state=None, 

shrinking=True, tol=0.001, verbose=False) 

   

 
Fig. 4. SVM Output results 

 
Fig. 4 elaborates that precision of 0.83 and 

0.57 have been achieved for the class 2 and 
class 4 respectively. 0.72 average precision has 
been achieved by using simple support vector 
machine.  

4.2. Kernel Support Vector Machine 
(SVM)  

Simple Support vector machine is applied 
to classify the linearly separable data. But for 
the classification of non-linear data straight 
line of decision boundary cannot be used. 
Therefore, a variant of support vector machine 
named as Kernel was proposed which included 
the mathematical solution as well for solving 
the non-linearly separable data. Scikit-learn 
was needed to implement the Kernel support 
vector machine with these parameters.  

SVC(C=1.0, cache_size=200, class_weight=

None, coef0=0.0, 

decision_function_shape='ovr', degree=8, ga

mma='auto_deprecated', 

kernel='poly', max_iter=-1, probability=False

, random_state=None, 

shrinking=True, tol=0.001, verbose=False) 

 

 
Fig. 5. Kernel SVM results 



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

5 

 

Fig. 4 shows that precision, recall. F1-score 

and support have been calculated to determine 

the performance of Kernel SVM classifier.  

4.3. Gaussian Kernel Support Vector 
Machine (SVM) 

k(x, x0) = exp ³−||x − x0||2/2σ2´                 (3)   

Gaussian Kernel support vector machine was 

applied to the breast cancer data set and 

following results were achieved.               

SVC(C=1.0, cache_size=200, class_weight=

None, coef0=0.0, 

 decision_function_shape='ovr', degree=3, ga

mma='auto_deprecated', 

kernel='rbf', max_iter=-1, probability=False, 

random_state=None, 

shrinking=True, tol=0.001, verbose=False) 

 
Fig. 6. Gaussian Kernel SVM results 

4.4. Sigmoid Kernel Support Vector 
Machine (SVM)  

Sigmoid Kernel can be applied to the data set 

for the classification by using the following 

mathematical equation 

          (4) 

SVC(C=1.0, cache_size=200, class_weight=

None, coef0=0.0, 

decision_function_shape='ovr', degree=3, ga

mma='auto_deprecated', 

kernel='sigmoid', max_iter=-1, probability=F

alse, random_state=None, 

shrinking=True, tol=0.001, verbose=False) 

 

 
Fig. 7. Sigmoid Kernel SVM results 

Recall is the measurement of corrected 

classified values out of the all positive classes. 

The higher the recall the better the 

performance. Recall can be measured by using 

eq. (10). 

 

𝑅𝑒𝑐𝑎𝑙𝑙 =
𝑇𝑝

𝑇𝑃+𝐹𝑁
                                   (5)                                        

 

F-measure can be acknowledged as the 

comparative analysis or to know the 

comparison between recall and precision. 

 

𝐹 − 𝑚𝑒𝑎𝑠𝑢𝑟𝑒 =
2×𝑅𝑒𝑐𝑎𝑙𝑙×𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛

𝑅𝑒𝑐𝑎𝑙𝑙+𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛
(6)                                                                                        

 

The estimation of actual positive out of all 

positive can be classes can be found be 

precision. 

 

𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 =
𝑇𝑃

𝑇𝑃+𝐹𝑃
                               (7)  

4.5. Fine Gaussian Support Vector 
Machine (SVM)  

 
     Fig. 8. Confusion Matrix of fine Gaussian 

SVM 

Confusion matrix of fine Gaussian SVM 

elaborated that the algorithm performed very 

https://d2h0cx97tjks2p.cloudfront.net/blogs/wp-content/uploads/sites/2/2017/08/sigmoid-kernel.png


 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

6 

 

poor for the prediction of breast cancer as it 

classified all classes as class 1. Class 2 was not 

predicted at all therefore false negative rate 

was found to be 100% and false positive rate 

for class 2 was found to be 0%.  

5. Comparative Analysis 

Table II: SVM Parametric Evaluation 

(CLASS 2) 
Classifiers Precision Recall F1-

Score 

Support 

SVM 0.83 0.45 0.59 11 
Kernel 

SVM 
0.0 0.0 0.0 12 

Sigmoid 

Kernel 

SVM 

0.60 1 0.75 12 

Gaussian 
Sigmoid 

Kernel 

0.60 1 0.75 12 

 

Table II demonstrated that the SVM and its 

variants’ parametric analysis for the 

classification of class 2. It can be observed that 

linear support vector machine achieved better 

results in terms of precision, recall, F1-score. 

 

Table III: SVM Parametric Evaluation 

(CLASS 4) 
Classifiers Precision Recall F1-

Score 

Support 

SVM 0.57 0.89 0.70 9 
Kernel 
SVM 

0.4 1 0.57 8 

Sigmoid 

Kernel 

SVM 

0 0 0 8 

Gaussian 

Sigmoid 

Kernel 

0 0 0 8 

 

Table III demonstrated that the SVM and its 

variants’ parametric analysis for the 

classification of class 4. It can be observed that 

linear support vector machine achieved better 

results in terms of precision, recall, F1-score. 

 

6. Results and discussion 

Support vector machine with its variants 
Kernel, Gaussian Kernel and Sigmoid Kernel 
have been applied to the data set of breast 
cancer. All the variants performed better. 0.57 
have been achieved for the class 2 and class 4 
respectively. 0.72 average precision has been 
achieved by using simple support vector 
machine. For the future work SVM can be 
combined with the Particle swarm 
optimization in which weights of Support 
vector machine would be optimized by PSO. 
The optimization of SVM weights would make 
the results better and up to the mark for the 
classification of breast cancer. 

REFERENCES 

[1] T. Botterill, T. Lotz, A. Kashif and J. G. 
Chase, "Reconstructing 3-D Skin Surface 
Motion for the DIET Breast Cancer 
Screening System," in IEEE 
Transactions on Medical Imaging, vol. 
33, no. 5, pp. 1109-1118, May 2014. 
doi: 10.1109/TMI.2014.2304959 

[2] L. He, D. Wang and Z. Guo, 
"Identification of Potential Non-invasive 
Biomarkers for Breast Cancer Prognosis 
and Treatment by Systematic 
Bioinformatics Analysis," 2015 7th 
International Conference on Information 
Technology in Medicine and Education 
(ITME), Huangshan, 2015, pp. 117-120. 
doi: 10.1109/ITME.2015.28 

[3] A. Helwan and R. H. Abiyev, "ISIBC: An 
intelligent system for identification of 
breast cancer," 2015 International 
Conference on Advances in Biomedical 
Engineering (ICABME), Beirut, 2015, pp. 
17-20. 
doi: 10.1109/ICABME.2015.7323240 

[4] Meharun Nisa S.P and K. Suresh, 
"Labview implementation of 
identification of early signs of breast 
cancer," 2014 International Conference 
on Electronics, Communication and 
Computational Engineering (ICECCE), 
Hosur, 2014, pp. 264-267. 
doi: 10.1109/ICECCE.2014.7086624 

[5] D. Agarwal, M. Kergosien, D. J. 
Boocock, R. C. Rees and G. R. Ball, "A 
systems biology approach to identify 
proliferative biomarkers and pathways in 
breast cancer," 2014 IEEE International 



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

7 

 

Conference on Bioinformatics and 
Biomedicine (BIBM), Belfast, 2014, pp. 
1-7. 
doi: 10.1109/BIBM.2014.6999240 

[6] N. Kiarashi et al., "Development and 
Application of a Suite of 4-D Virtual 
Breast Phantoms for Optimization and 
Evaluation of Breast Imaging Systems," 
in IEEE Transactions on Medical 
Imaging, vol. 33, no. 7, pp. 1401-1409, 
July 2014. 
doi: 10.1109/TMI.2014.2312733 

[7] P. Swapnil, E. Pandey, J. R. Yathav, A. 
Baig and A. Bailur, "Region Marking and 
Grid Based Textural Analysis for Early 
Identification of Breast Cancer in Digital 
Mammography," 2016 IEEE 6th 
International Conference on Advanced 
Computing (IACC), Bhimavaram, 2016, 
pp. 426-429. 
doi: 10.1109/IACC.2016.85 

[8] J. Kamalakannan, P. V. Krishna, M. R. 
Babu and K. D. Mukeshbhai, 
"Identification of abnormility from digital 
mammogram to detect breast 
cancer," 2015 International Conference 
on Circuits, Power and Computing 
Technologies [ICCPCT-2015], 
Nagercoil, 2015, pp. 1-5. 
doi: 10.1109/ICCPCT.2015.7159454 

[9] E. B. Nababan, M. Iqbal and R. F. 
Rahmat, "Breast cancer identification on 
digital mammogram using Evolving 
Connectionist Systems," 2016 
International Conference on Informatics 
and Computing (ICIC), Mataram, 2016, 
pp. 132-136. 
doi: 10.1109/IAC.2016.7905703 

[10] S. Pathan, P. C. Siddalingaswamy, L. 
Lakshmi and K. G. Prabhu, 
"Classification of benign and malignant 
melanocytic lesions: A CAD tool," 2017 
International Conference on Advances in 
Computing, Communications and 
Informatics (ICACCI), Udupi, 2017, pp. 
1308-1312. 
doi: 10.1109/ICACCI.2017.8126022 

[11] I. E. Magnin, D. Vray and A. Brémond, 
"Early detection of breast cancer using 
computer assisted diagnosis," 1992 14th 
Annual International Conference of the 
IEEE Engineering in Medicine and 
Biology Society, Paris, 1992, pp. 849-850.  

[12] R. D. Ghongade and D. G. Wakde, 
"Detection and classification of breast 
cancer from digital mammograms using 
RF and RF-ELM algorithm," 2017 1st 
International Conference on Electronics, 

Materials Engineering and Nano-
Technology (IEMENTech), Kolkata, 
2017, pp. 1-6. 
doi: 
10.1109/IEMENTECH.2017.8076982 

[13] P. Král and L. Lenc, "LBP features for 
breast cancer detection," 2016 IEEE 
International Conference on Image 
Processing (ICIP), Phoenix, AZ, 2016, 
pp. 2643-2647. 
doi: 10.1109/ICIP.2016.7532838  

[14] B. Hela, M. Hela, H. Kamel, B. Sana and 
M. Najla, "Breast cancer detection: A 
review on mammograms analysis 
techniques," 10th International Multi-
Conferences on Systems, Signals & 
Devices 2013 (SSD13), Hammamet, 
2013, pp. 1-6. 
doi: 10.1109/SSD.2013.6563999  

[15] T. Kao et al., "Regional Admittivity 
Spectra With Tomosynthesis Images for 
Breast Cancer Detection: Preliminary 
Patient Study," in IEEE Transactions on 
Medical Imaging, vol. 27, no. 12, pp. 
1762-1768, Dec. 2008. 

[16] D. A. Woten, J. Lusth and M. El-
Shenawee, "Interpreting Artificial Neural 
Networks for Microwave Detection of 
Breast Cancer," in IEEE Microwave and 
Wireless Components Letters, vol. 17, no. 
12, pp. 825-827, Dec. 2007. 
doi: 10.1109/LMWC.2007.910466 

[17] P. M. Meaney, M. W. Fanning, D. Li, S. 
P. Poplack, and K. D. Paulsen,“A clinical 
prototype for active microwave imaging 
of the breast,”IEEE Trans. Microw. 
Theory Tech., vol. 48, no. 11, pp. 1841–
1853,Nov. 2000doi: 
10.1109/MCS.2009.932223 

[18] Wolberg, W.H., & Mangasarian, O.L. 
(1990). Multisurface method of pattern 
separation for medical diagnosis applied 
to breast cytology. In Proceedings of the 
National Academy of Sciences, 87, 9193-
-9196. 

 

[19] Zhang, J. (1992). Selecting typical 
instances in instance-based learning. In 
Proceedings of the Ninth International 
Machine Learning Conference (pp. 470--
479). Aberdeen, Scotland: Morgan 

[20] T. Khan et al., "Foreign objects debris 
(FOD) identification: A cost effective 
investigation of FOD with less false 
alarm rate," 2017 IEEE 4th International 
Conference on Smart Instrumentation, 



 
Talha Ahmed Khan(et al.), Performance Comparison of SVM and its Variants for the early prognosis of Breast cancer                                   

(pp. 1 - 8) 

Sukkur IBA Journal of Computing and Mathematical Science - SJCMS | Vol. 3 No. 2 July - December 2019 © Sukkur IBA University 

8 

 

Measurement and Application 
(ICSIMA), Putrajaya, 2017, pp. 1-4. 

[21] Talha Ahmed Khan, Kushsairy Kadir, 
Muhammad Alam, Zeeshan Shahid and 
M.S. Mazliham, “Geomagnetic Field 
Measurement at Earth Surface: Flash 
Flood Forecasting using Tesla Meter”, 
Proc. of the International Conference on 
Engineering Technologies and 
Technopreneurship (IEEE-ICE2T 2017) 
18-20 September 2017, Kuala Lumpur, 
Malaysia 

[22] T. A. Khan, M. Alam, K. Kadir, Z. Shahid 
and S. M Mazliham, "A Novel Approach 
for the Investigation of Flash Floods 
using Soil Flux and CO2: An 
Implementation of MLP with Less False 
Alarm Rate," 2018 2nd International 
Conference on Smart Sensors and 
Application (ICSSA), TBA, Malaysia, 
2018, pp. 130-134 

[23] Talha Khan, Muhammad Alam, 
Kushsairy Kadir, Sheroz Khan, M.S 
Mazliham, Faraz Shaikh, Syed Faiz 
Ahmed, Zeeshan Shahid” An 
Implementation of Electroencephalogram 
signals acquisition to control manipulator 
through Brain Computer Interface “, 2nd 
IEEE International Conference on 
Innovative research and development 
2019 (IEEE-ICIRD), 28-30 June 2019, 
Jakarta, Indonesia