Adapting God Class thresholds for software defect prediction: A case study

Gradisnik, Mitja ✉; Beranic, Tina; Karakatic, Saso; Mausa, Goran

English Scientific
    In software engineering there is an active research field of defect prediction using software metrics. While the research shows that the prediction of defects using software metrics performs well, prediction using metrics alone lacks clear refactoring capabilities. On the other hand, code smells have the ability to describe the code anomalies precisely, and suggest their refactoring. Therefore, code smells can be a much better starting position for software fault prediction.In this paper, we present the results of preliminary research on the ability to predict software defects with the code smell God Class. The aim of our research was to test the definition of God Class, as defined by Lanza and Marinescu in 2006, in the ability to predict defects in a case study of the open source projects JDT and PDE within the Eclipse framework. The definition of the God Class was adapted using the grid search technique, with the goal of maximizing the fault prediction ability while keeping the base of the original definition. The results show that adaption of the definition in the specific project resulted in improved fault prediction ability.
    Citation styles: IEEEACMAPAChicagoHarvardCSLCopyPrint
    2021-09-23 17:55