SVM with feature selection and extraction techniques for defect-prone software module prediction

Abstract

In this paper, support vector machines with combinations of different feature selection and extraction techniques are used for the prediction of defective software module. It is tested on five NASA datasets. Correlation-based feature selection technique (CFS), principal component analysis (PCA) and kernel principal component analysis (KPCA or kernel PCA) techniques are used for feature selection and feature extraction. It has been shown that the CFS + SVM gives better prediction results and accuracy compare to PCA + SVM and KPCA + SVM.