A Sequential Comparative Analysis of Software Change Proneness Prediction Using Machine Learning

Abstract

Change-prone modules are more likely to produce defects and accumulate technical debt. Thus, developing prediction models for determining change-prone software classes is critical. Such models will allow for more efficient resource utilization during the maintenance phase and will make them more adaptable to future changes. This paper applies the study on a large dataset from a commercial software to investigate the relationships between object-oriented metrics and change-proneness. The study also compared the performance of several machine learning techniques including combining methods that were constructed by combining several single and ensemble classifiers with voting, Select-Best, and stacking scheme. The result of the study indicates a high prediction performance of many of the ensemble classifiers and the combining methods selected and proved that machine learning methods are very beneficial for predicting change-prone classes in software. The study also demonstrated that software metrics are significant indicators of class change-proneness and should be monitored regularly.