Cross‐project defect prediction method based on genetic algorithm feature selection

With the continuous development of Internet technology, role software in life is increasing, and defect prediction (SDP) a key means to ensure reliability. SDP predict modules that may have defects advance based on historical data projects, its purpose maximize use testing resources. However, actual process, project needs be predicted often new for which there little or no data. Therefore, how massive other related projects build cross-project (CPDP) model has received extensive attention from scholars. due differences distribution class imbalance between different performance CPDP greatly affected. basis CPDP, this article proposes feature selection method genetic algorithm (genetic selection, GAFS). GAFS mainly includes two stages: ensemble training. In stage, global search adaptive algorithm, uses integrated training results candidate subsets target migrate optimal subset. phase, EasyEnsemble used alleviate problem, multiple naive Bayesian classifiers are constructed, then final constructed through learning. article, F1-score MCC as test indicator, comparative experiments carried out AEEEM Promise. The show compared with five comparison methods, can improve average much more. For example, value by 38.9%, 31.6%, 35.1%, 22.0%, respectively. most cases, it effectively achieve better results.