Improving the performance of many-objective software refactoring technique using dimensionality reduction

Abstract

Software quality Assessment involves the measurement of a large number of software attributes referred to as quality metrics. In most searchedbased software engineering processes, an optimization algorithm is used to evaluate a certain number of maintenance operations by minimizing or maximizing these quality metrics. One such process is software refactoring. When the solution to the problem includes a large number of objectives, various difficulties arise, including the determination of the Pareto-optimal front, and the visualization of the solutions. However, in some refactoring problem, there may be redundancies among any two or more objectives. In this paper, we propose a new software refactoring approach named PCA-NSGA-II many-objective refactoring. This approach is based on the PCA-NSGA-II evolutionary multi-objective algorithm, and can overcome the curse of dimensionality by removing redundancies to retain conflicting objectives for further analysis.