An Enhanced Evolutionary Software Defect Prediction Method Using Island Moth Flame Optimization

Software defect prediction (SDP) is crucial in the early stages of defect-free software development before testing operations take place. Effective SDP can help test managers locate defects and defect-prone modules. This facilitates allocation limited quality assurance resources optimally economically. Feature selection (FS) a complicated problem with polynomial time complexity. For dataset N features, complete search space has 2N feature subsets, which means that algorithm needs an exponential running to traverse all these subsets. Swarm intelligence algorithms have shown impressive performance mitigating FS reducing time. The moth flame optimization (MFO) well-known swarm been used widely proven its capability solving various problems. An efficient binary variant MFO (BMFO) proposed this paper by using island BMFO (IsBMFO) model. IsBMFO divides solutions population into set sub-populations named islands. Each treated independently BMFO. To increase diversification algorithm, migration step performed after specific number iterations exchange between Twenty-one public datasets are for evaluating method. results experiments show improves classification results. followed support vector machine (SVM) best model over other compared models, average G-mean 78%.