Machine Learning Implementation for Refactoring Prediction

Abstract

Refactorings improve the internal organization of object-oriented software project without altering the functionality to address the problem of architectural degradation. The application of refactoring leads to increased software quality and maintainability. However, finding refactoring chances is a complex topic that affects both developers and researchers. In a recent study, machine learning methods demonstrated significant promise for resolving this issue. Model refactoring prevents erosion of the program architecture at an early stage of the model-driven engineering paradigm-compliant software development project. However, difficulties such as variable data set distribution and the availability of duplicate and irrelevant variables hamper the efficacy of refactoring prediction models. We aim to develop a model for refactoring prediction using several machine learning classifiers, data sampling techniques, and feature selection techniques.