Cross-Project Software Defect Prediction Based on Feature Selection and Knowledge Distillation

Cross-project software defect prediction (CPDP) utilizes labelled data from source projects to predict defects in target projects, aiding engineers in defect detection and resolution. Nonetheless, the original defect dataset often contains a large number of redundant and irrelevant features. Moreover, most feature selection methods are limited to feature selection on a single project and fail to simultaneously filter the features of source and target projects. Furthermore, distributional differences between source and target projects result in poor prediction outcomes when directly applying models trained on source projects to target projects. These challenges hinder the effectiveness of CPDP. In this paper, we introduce a cross-project software defect prediction model based on feature selection and knowledge distillation (FSKDP) to overcome these limitations.