Volume 5, Issue 2 (Summer 2018)
jhbmi 2018, 5(2): 244-251 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Orooji A, Langarizadeh M. Evaluation of the Effect of Feature Selection and Different kernel Functions on SVM Performance for Breast Cancer Diagnosis. jhbmi 2018; 5 (2) :244-251
URL: http://jhbmi.ir/article-1-284-en.html
URL: http://jhbmi.ir/article-1-284-en.html
PhD of Medical Informatics, Health Information Management Dept., School of Health Management and Information Sciences, Iran University of Medical Sciences, Tehran, Iran.
Abstract: (6834 Views)
Introduction: Breast cancer is one of the most common cancers affecting women. In mammography, differentiating a malignant tumor from a benign one is a very tedious task due to their structural similarities. Machine Learning (ML) is a subfield of Artificial Intelligence that can be used as an effective tool to help physicians to make decisions. Support vector machine (SVM) is one of the most common ML techniques that its performance depends on kernel parameters tuning and input features. The aim of this study was to investigate the effect of feature selection and different kernel functions on SVM performance.
Methods: This analytic study was performed through comparative method. Genetic algorithm was used for feature selection. SVM models based on different kernel functions, including polynomial, Linear, Radial Basis Function (RBF), Quadratic and Multi-Layer Perceptron (MLP), were first performed with all features and then, with the selected features. The Wisconsin original breast cancer data set was used as a training set to evaluate the performance of the classifiers. All implementations were done in MATLAB environment.
Results: According to the obtained results, by applying feature selection, the performance of SVM with MLP kernel function decreased and with quadratic kernel function increased. However, the performances of the linear and RBF kernels were desirable in both conditions. Generally, after the dimension reduction, the best accuracy, specificity, sensitivity and accuracy were dropped by 0.663, 0.833, 1.077 and 0.138 percent respectively.
Conclusion: The ML-based methods can help physicians in diagnosis and decision makings for treatment.
Methods: This analytic study was performed through comparative method. Genetic algorithm was used for feature selection. SVM models based on different kernel functions, including polynomial, Linear, Radial Basis Function (RBF), Quadratic and Multi-Layer Perceptron (MLP), were first performed with all features and then, with the selected features. The Wisconsin original breast cancer data set was used as a training set to evaluate the performance of the classifiers. All implementations were done in MATLAB environment.
Results: According to the obtained results, by applying feature selection, the performance of SVM with MLP kernel function decreased and with quadratic kernel function increased. However, the performances of the linear and RBF kernels were desirable in both conditions. Generally, after the dimension reduction, the best accuracy, specificity, sensitivity and accuracy were dropped by 0.663, 0.833, 1.077 and 0.138 percent respectively.
Conclusion: The ML-based methods can help physicians in diagnosis and decision makings for treatment.
Keywords: Machine Learning, Dimension reduction, Clinical decision support systems, Early cancer diagnosis
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
