Comprehensive Examination of Machine Learning Methods for Predicting Software Errors

This paper presents a comprehensive examination of machine learning methods for predicting software errors, emphasizing the importance of algorithm selection, data quality, and class imbalance in the development process. The study analyzes various machine learning classifiers— including Support Vector Machine (SVM), C4.5, Logistic Regression (LR), Ripper, Instance-Based learner (IBk), Random Forest (RF), and Naive Bayes (NB)—and evaluates their performance under varying conditions of data imbalance, measured by the Matthews Correlation Coefficient (MCC). The findings reveal significant variability in classifier performance, highlighting the challenges associated with predictive modeling in imbalanced datasets. The research contributes to the field of software engineering by proposing a framework for integrating machine learning into software development workflows, aiming to enhance software reliability and mitigate the impact of errors in production environments.