Comparative Analysis of Prediction Models for Software Bug Prediction

This paper explores the prediction of software bugs at varying levels of granularity, evaluating different prediction models in this context. The study incorporates 18 open-source datasets for method-level bug predictions, incorporating code metrics and historical measures. Various models like Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Naive Bayes (NB), and Decision tree (DT) are evaluated using metrics like F1 Score, AUC-ROC, Matthews Correlation Coefficient MCC, Precision, G-mean, and Accuracy. While Random Forest consistently performs well in AUC, some models excel across metrics and datasets. The findings underscore the importance of model selection based on dataset characteristics. This crucial aspect of our research provides a solid foundation for software developers to confidently choose the most suitable bug prediction models for their specific datasets, leading to more efficient and reliable software development practices.