ANALYSIS OF MACHINE LEARNING METHODS TO DETECT SOFTWARE DEFECTS

Vishnevskaya Tatyana Ivanovna, Klimov Ilya Sergeevich

This article provides an overview of software defects that occur both at the development stage and in the process of using the software product. There are three groups in the classification of methods for their detection: static, dynamic and operational. The possibility of using machine learning to detect defects in the software being developed is considered. An overview of the most common machine learning methods (naive Bayesian classifier, support vector machine, decision tree, random forest, boosting) and the results of comparing these methods based on selected metrics are presented. In conclusion, it is concluded that the gradient boosting method is the most effective and promising for the problem of detecting software defects.

Keywords:defect, machine learning, detection, methods