Methods of predicting defects in software products based on retrospective and current information
Abstract
This paper examines the management problems in organizational systems of IT companies related to predicting defects in the source code of software products. The purpose of this work is to determine the most effective method for accurately predicting defects in the source code of software products implemented by organizational systems of IT companies. The relevance of the work is due to the fact that traditional statistical methods and models do not provide a sufficiently accurate forecast of the quality of the product, and static code analysis tools have a large number of false positives and false negatives. In this regard, there is an increasing need to apply new methods in predicting errors in the code of software products. At the same time, it is necessary to take into account that not all new algorithms and models implementing the methods are suitable for solving this problem. The object of the study is the growing demand for high-quality software products implemented by organizational systems of IT companies. The subject of the study is the metrics for assessing the quality of methods and algorithms designed to predict defects in the source code of software products. The performance evaluation metrics are accuracy, completeness and F1-measure. As a result of the experiments, conclusions were made about the performance of each model, and the choice of the most suitable algorithm for solving this applied problem was determined. The prospects for further research aimed at improving the performance of defect prediction methods in the source code of software products implemented by the organizational systems of IT companies are outlined. The methodological basis of the work is the methods of analysis, comparison, experiment and generalization.