Software Refactoring Using Cooperative Parallel Evolutionary Algorithms

Code bad smells represent symptoms of poor implementation choices. Previous studies found that these smells make source code more difficult to maintain, possibly also increasing its fault-proneness. There are plethora of approaches that identify bad smells based on code analysis techniques. We propose, in this paper, to consider code-smells correction as a distributed problem. Different techniques are combined during an optimization process to find a consensus regarding the correction of code-smells. To this end, we used Distributed Evolutionary algorithms (D-EA) where many evolutionary algorithms with different adaptations (fitness functions, solution representation, and change operators) are executed in parallel to solve a common goal which is the correction of code-smells. The statistical analysis of the obtained results provide evidence to support the claim that cooperative D-EA outperforms single population evolution and random search based on a benchmark of eight large open source systems where more than 86% of codesmells are fixed using the suggested refactorings. Keywords—Search-based software engineering, applications, software refactoring, evolutionary algorithms.