Cross-Project Defect Prediction with Metrics Selection and Balancing Approach

Abstract In software development, defects influence the quality and cost in an undesirable way. Software defect prediction (SDP) is one of techniques which improves testing efficiency by early identification defects(bug/fault/error). Thus, several experiments have been suggested for (DP) techniques. Mainly DP method utilises historical project data constructing models. SDP performs well within projects until there adequate amount accessible to train However, if are inadequate or limited same project, researchers mainly use Cross-Project Defect Prediction (CPDP). CPDP a possible alternative option that refers anticipating using models built on from other projects. challenging due its distribution domain difference problem. The proposed framework effective two-stage approach CPDP, i.e., model generation process. phase, conglomeration different pre-processing, including feature selection class reweights technique, used improve initial quality. Finally, fine-tuned efficient bagging boosting based hybrid ensemble developed, avoids over -fitting/under-fitting helps enhance performance. process generated predicts projects, has clean. evaluated using25 obtained public repositories. result analysis shows achieved 0.71±0.03 f1-score, significantly state-of-the-art approaches 23 % 60 %.