🐞 BugClassify: Semantic Bug Report Classification

A multi-stage deep learning framework to automatically classify software bug reports with high accuracy using state-of-the-art NLP.

---

🚀 Features

• 🧠 State-of-the-Art Semantic Understanding — Utilizes a fine-tuned Sentence-BERT (SBERT) model to understand the true meaning of bug reports, not just keywords.
• 🏗️ Specialized Dual-Model Architecture — Employs a lightweight 1D CNN for fast initial triage and a powerful Transformer for deep, nuanced analysis.
• 💡 Innovative Data Augmentation — Introduces domain-specific hard-negative sampling to create a robust model that isn't fooled by technical jargon.
• 🔬 Multi-Level Classification — Categorizes reports by Existence (Bug/Non-Bug), Complexity (Bohrbug/Mandelbug), and Root Cause (ARB/NAM).
• 📊 Proven High Performance — Achieves an F1-Score of 0.9644 on highly imbalanced, real-world data where traditional methods fail.
• 🔌 Modular & Extensible — Designed to be easily integrated into existing bug tracking workflows and extended for other software engineering tasks.

---

📦 Installation

1. Clone the repository:

   git clone https://github.com/Samurai007AK/BUgClassify.git
   cd BugClassify

2. Install the required packages:

   pip install -r requirements.txt

---

🛠️ How to Use

(This is a template. You should update this with your actual script names and arguments.)

1. Train the semantic model:

   python train_semantic_model.py --data_path ./data --output_dir ./models/sem_model_4

2. Run the Bug vs. Non-Bug classification:

   python run_classification.py --model_type cnn --data_path ./data/linux_dataset.csv --semantic_model ./models/sem_model_4

3. Run the advanced classification:

   python run_classification.py --model_type transformer --data_path ./data/curated_linux.xlsx --semantic_model ./models/sem_model_4

📂 Repository Structure

BugClassify/
│
├── data/                     # Folder for datasets (e.g., Linux, GCC, LLVM)
├── models/                   # Folder to save trained models and embeddings
├── notebooks/                # Jupyter notebooks for exploration and analysis
├── src/                      # Source code for the project
│   ├── preprocessing.py      # Data cleaning and augmentation scripts
│   ├── models.py             # Definitions for the CNN and Transformer architectures
│   ├── train.py              # Script for training the models
│   └── evaluate.py           # Script for evaluating model performance
│
├── requirements.txt          # List of Python dependencies
└── README.md                 # This file

---