BMSFP: Best Metric Suite for Software Fault Prediction Based on Feature Selection Technique

Software error forecasting, a critical aspect of software engineering, aims to identify faultful software system modules to enhance software quality and reliability. The effectiveness of fault predictions hinges on the selection of metrics and feature selection techniques. This paper critically reviews the most influential metrics and feature selection techniques used in software fault prediction, evaluates their effectiveness, and identifies the best metric suites to date. By amalgamating recent research findings, the study not only underscores the potential of these approaches but also highlights areas where further solidification could lead to more profound insights. We have compared the different classifiers based on NASA's KC1 dataset. The result shows that Random Forest has the best accuracy. We stress the need for continued investigation and improvement in software fault prediction techniques, and our study provides a solid foundation for such endeavors.