Comparison of software defect prediction models based on machine learning

Abstract

Software defect prediction has been widely used in software system development, among which the method based on machine learning has proved to be more effective. Firstly, the basic framework of the prediction model and the metric elements used in the prediction process are introduced in this paper. Secondly, the three main machine learning-based software defect prediction models (LR, SVM, and BPNN) are analyzed, and finally the prediction effects of the three models are compared and analyzed by using the experimental results of MDP.