This project explores the automation of bug report classification and assignment using Large Language Models (LLMs) and Machine Learning (ML) techniques. By leveraging LLM-based ALBERT embeddings for feature extraction and combining them with traditional classifiers like Random Forest and Support Vector Machine (SVM), the study aims to enhance the efficiency of bug triaging in software maintenance.
- ✅ LLM-Based ALBERT for Feature Extraction: Utilizes pre-trained transformer-based embeddings to capture semantic features from bug reports.
- ✅ Machine Learning Models: Implements Random Forest, SVM, and Grid Search Optimized RF for classification.
- ✅ Imbalanced Data Handling: Analyzes performance across major and minor bug severity classes.
- ✅ Performance Evaluation: Compares models using Precision, Recall, and F1-score metrics.
- ✅ Efficient Local Execution: Conducted experiments on a 16GB RAM, Intel Core i7-6700HQ system using Python 3.12 and Jupyter Notebook.
- Python 3.12
- Large Language Models (LLMs)
- ALBERT (A Lite BERT for Self-supervised Learning of Language Representations)
- Random Forest, SVM, and Grid Search CV
- NumPy, Pandas, Scikit-Learn
- 📌 LLM-Based ALBERT + Random Forest demonstrated superior accuracy in handling major bug classes.
- 📌 Class imbalance remained a challenge, highlighting the need for data augmentation or custom loss functions.
- 📌 Optimized models showed improved precision-recall trade-offs, aiding in effective bug assignment.
- 🔹 Incorporate deep learning-based classifiers for further performance improvements.
- 🔹 Explore synthetic data generation to handle class imbalance.
- 🔹 Develop a real-time bug triaging system for industry applications.
- Python 3.x
- pip (Python package installer)
- Clone the repository:
git clone https://github.com/Baybordi/Bug-Report-Classification-ALBERT