유방암 분류 성능 향상을 위한 배깅 서포트 벡터 머신

Abstract

Objectives: We proposed bagging SVM which constructs SVM ensembles using bagging for improving breast cancer classification. Methods: Each individual SVM was trained independently using the randomly chosen training samples via a bootstrap technique. Then, they were aggregated into to make a collective decision in aggregation strategy such as the majority voting. We compared the proposed bagging SVM model with existing single models such as discriminant analysis, logistic regression analysis, decision tree, support vector machines for two UCI data and simulated data. Performance of these techniques was compared through accuracy, positive predictive value, negative predictive value, sensitivity, specificity and F-score. Results: Experimental results for two UCI data and the simulated data showed that the proposed bagging SVM model outperformed single SVM, discriminant analysis, logistic regression analysis, decision tree and neural network in terms of various performance measures. Conclusions: We proposed bagging SVM for improving breast cancer classification. The bagging SVM ensembles outperformed existing single models for all applications in terms of various performance measures. Keywords: Classification, Breast cancer, Support vector machine, Performance evaluation, Bagging support vector machine