Designing a Robust Software Bug Prediction Model Using Enhanced Learning Principles with Artificial Intelligence Assistance

Software bug prediction is a critical task in software engineering, aimed at identifying bug-prone areas in the codebase before they cause significant issues. In this study, we propose a robust hybrid model that integrates Long Short-Term Memory (LSTM) networks and eXtreme Gradient Boosting (XGBoost) to enhance bug prediction accuracy. The proposed model leverages the sequential learning capabilities of LSTM to capture temporal patterns in software development, while XGBoost handles high-dimensional feature spaces for accurate classification. The model was trained and tested using historical software bug data collected from open-source repositories, with features including source code metrics, code review history, and developer activity. The proposed model achieved a remarkable accuracy of 92.81 %, outperforming nine existing models, including Random Forest, Support Vector Machine (SVM), and Artificial Neural Networks (ANN). Additionally, the model demonstrated superior performance in recall (91.43%) and F1-score (0.915), indicating its effectiveness in both identifying bugs and minimizing false positives. These results suggest that the proposed hybrid model is a powerful tool for proactive bug detection, offering significant improvements over traditional methods.