Improved Forecasting of Software Defect: A Cross Analysis of Machine Learning Approaches with Champ Feature Optimization

Predicting software defects is vital for maintaining software reliability, as it helps in identifying faulty components early in the development phase. This work looks at how selecting the right features affects the accuracy of predictions, using the CHAMP optimization technique to find important features from NASA datasets like CM1, KC2, MC1, and PC1. The study uses three models-support vector machine (SVM), extreme learning machine (ELM), and random forest (RF)-to predict defects and compares the results of predictions made with specific features chosen and those made without them. The results show that incorporating feature selection significantly boosts accuracy, with ELM achieving the highest performance on the KC2 dataset when using features selected through CHAMP. These results highlight how important it is to choose the right features to improve software defect prediction (SDP) and show that using ELM with CHAMP can result in more accurate predictions.