On the applicability of evolutionary computation for software defect prediction

Removal of defects is the key in ensuring long-term error free operation of a software system. Although improvements in the software testing process has resulted in better coverage, it is evident that some parts of a software system tend to be more defect prone than the other parts and identification of these parts can greatly benefit the software practitioners in order to deliver high quality maintainable software products. A defect prediction model is built by training a learner using the software metrics. These models can later be used to predict defective classes in a software system. Many studies have been conducted in the past for predicting defective classes in the early phases of the software development. However, the evolutionary computation techniques have not yet been explored for predicting defective classes. The nature of evolutionary computation techniques makes them better suited to the software engineering problems. In this study we explore the predictive ability of the evolutionary computation and hybridized evolutionary computation techniques for defect prediction. This work contributes to the literature by examining the effectiveness of the 15 evolutionary computation and hybridized evolutionary computation techniques to 5 datasets obtained from the Apache Software Foundation using the Defect Collection and Reporting System. The results are evaluated in terms of the values of accuracy. We further compare the evolutionary computation techniques using the Friedman ranking. The results suggest that the defect prediction models built using the evolutionary computation techniques perform well over all the datasets in terms of prediction accuracy.