Software fault prediction using a differential evolution-based wrapper approach for feature selection

Software fault prediction (SFP) aids in the early sensing of software flaws, hence improving the quality of software. The prediction process utilizes previously used software measurements and faulty data as autonomous characteristics to determine whether or not there is an issue in the software. Aboriginal detection of faults in software saves a significant amount of money space and labor costs which is necessary to remedy those flaws. However, because the magnitude of data is so large, feature selection(FS) is critical in order to obtain relevant information. This work provides an FS perspective based on the Differential Evolutionary Algorithm(DE) that uses the difference between pairs of randomly selected objective vectors for the mutation process in contrast to traditional Evolutionary Algorithms(EA) that rely on the preset probability distribution function for the same. For our tests, we employed a mix of DE, NBC, KNN, and DTC Classifiers. Our findings indicate that applying the DE for FS enhanced our accuracy in all three classifiers for the SFP data.