engleski

Software Defect Classification with a Variant of NSGA-II and Simple Voting Strategies

Software Defect Prediction is based on datasets that are imbalanced and therefore limit the use of machine learning based classification. Ensembles of genetic classifiers indicate good performance and provide a promising solution to this problem. To further examine this solution, we performed additional experiments in that direction. In this paper we report preliminary results obtained by using a Matlab variant of NSGA-II in combination with four simple voting strategies on three subsequent releases of the Eclipse Plug-in Development Environment (PDE) project. Preliminary results indicate that the voting procedure might influence software defect prediction performances.

SDP; SBSE; Multi-objective optimisation; NSGA-II

nije evidentirano