Advanced Bug Report Prediction Through Nature-Based Ensemble Machine Learning

In today’s software development world, quickly and accurately classifying bug reports is crucial for fixing issues on time and keeping software quality high. Bugs are a vital part of this process, carrying detailed information such as description, status, reporter, assignee, priority, severity, among others. It becomes tiresome and time-consuming to analyze these reports manually in search of all defects, especially when the size of these reports increases. Thus, automation is one of the best solutions. While current research is focused mainly on automating the identification of bug severity or priority, they often forget identification based on the nature of the bugs, which requires multi-class classification. A new prediction model has been used that analyzes BRs for the prediction of nature of the bugs. This model will combine the ensemble ML algorithm with NLP and ML techniques. In particular, the model is tested against publicly available datasets from Mozilla and Eclipse, which further categorize the bugs into six types: program anomaly, GUI, network or security, configuration, performance, and test-code. The results indicate that our model gave significantly higher accuracy compared to many pre-existing models without text augmentation, at 90.42 percent and with text augmentation at 96.72 percent.