Software Defect Prediction Model Based On KPCA-SVM

Abstract

Software defect prediction technology mainly relies on machine learning algorithm to learn the measurement data of existing software. There is some redundant data in the measurement element of software defect, which will reduce the accuracy of machine learning algorithm. This paper proposes a software defect prediction model based on KPCA-SVM.First, the dimension reduction pretreatment of software defect data sets is carried out.Then, This paper using support vector machines for classification.The accuracy of the model can be improved by keeping global features in the selection of the dimension reduction algorithm.Therefore, the kernel principal component analysis (KPCA) algorithm was selected for dimensionality reduction. For the selection of classification algorithm, this paper considering that the defect prediction data set has small samples and non-linear characteristics, the support vector machine has better advantages in this kind of data set, so SVM is selected as the classifier.In order to verify the performance of this model, this paper adopts the NASA MDP data set which is widely used in the field of software defect prediction.This paper use the CM1, JM1, PC1 and KC1 dataset to contrast KPCA -SVM model with a single SVM and LLE - SVM. This paper proved that KPCA - SVM model can better solve the problem of data redundancy of defect prediction data set.it can keep the global characteristics, and can have better prediction precision.