Determination of optimum refactoring sequence for maximizing the maintainability of object-oriented systems using machine learning algorithms

Abstract

Refactoring is a technique for changing internal attributes without affecting external ones in an optimized manner. Bad smells in the source code can cause various issues, increasing the need for refactoring. In this study, prioritization of classes is initially performed using a newly proposed metric called the Quality Decline Factor (QDF), which considers an appropriate ratio of software metrics along with eleven detected types of bad smells. Next, these bad smells are addressed by applying refactoring techniques, and changes in the metrics are observed. Subsequently, machine learning algorithms are used to assign weights to each metric, leading to the proposal of another new metric, the Total Refactoring Index (TRI). TRI combines the assigned weights and the effects of metric changes to determine the optimal refactoring sequence. The results indicate that the Decision Tree Forest algorithm is the most suitable for determining the refactoring sequence. After applying this technique, a 94.9% reduction in effort was observed. This study would benefit software maintainers by providing predefined sequences, allowing them to focus only on the code sections with the highest concentration of bad smells, thus completing projects within real-time constraints.