In-depth exploration of software defects and self-admitted technical debt through cutting-edge deep learning techniques
This study introduces an innovative deep learning–based approach to detect and classify Self-Admitted Technical Debt (SATD) along with related software defects. The model was trained using data from open-source projects such as Apache, Mozilla Firefox, and Eclipse, applying several architectures including LSTM, GRU, BERT, and GPT-3. Results show that the GPT-3 model achieved the highest accuracy (0.984), outperforming other models. The research contributes to improving software quality by enhancing the detection and understanding of technical debt and software defects, thereby supporting sustainable software maintenance and development. المزيد