🛠️ DeepTrace AI: A Deep Learning Framework for Multi-Domain Software Exception Classification & Automated Remediation

DeepTrace AI is an advanced academic and enterprise-grade Deep Learning project designed to automatically classify programming errors, compiler exceptions, and stack tracebacks across 7 core software domains, providing instant root-cause analysis and actionable fix recommendations.

---

🎓 Academic Project Overview

• Project Title: DeepTrace AI: A Deep Learning Framework for Multi-Domain Software Exception Classification and Automated Remediation
• Domain: Artificial Intelligence / Deep Learning / Natural Language Processing (NLP)
• Neural Architecture: Bidirectional Long Short-Term Memory (Bi-LSTM) + Word2Vec Continuous Skip-Gram Embeddings (100D) + Phrase N-Gram Mining
• Knowledge Engine: Contextual Regex Pattern Matching Knowledge Base
• Interface: Streamlit Interactive Diagnostic Studio

---

🎯 Supported Error Domains (7 Classes)

1. Database: PostgreSQL, MySQL, SQLite, MongoDB, Redis, ORMs (SQLAlchemy, Hibernate, Mongoose, Django DB).
2. Network: Connection timeouts, DNS failures, CORS policy blocks, SSL certificate errors, ECONNREFUSED.
3. API: HTTP 401/403/404/429/50x status codes, Axios/Fetch errors, GraphQL field errors, JSON format failures.
4. Dependency: Pip, NPM, Maven, Gradle, Cargo, Go Modules package missing & version resolution conflicts.
5. Syntax: Invalid syntax, indentation errors, missing semicolons/brackets, unclosed string literals.
6. Runtime: NullPointerException, IndexError, KeyError, ZeroDivisionError, OutOfMemoryError, Recursion.
7. Type: Unsupported operand types, incompatible type conversions, method signature mismatches, TS type errors.

---

⚙️ System Architecture Pipeline

graph TD
    A["Raw Error Message / Traceback Log"] --> B["NLP Preprocessing (CamelCase Split, Lowercase, Porter Stemming)"]
    B --> C["Phrase Mining (Word2Vec 100D Vector Space Embedding)"]
    C --> D["Neural Classifier Ensemble (Bi-LSTM / LSTM / CNN)"]
    D --> E["Predicted Category & Softmax Probabilities"]
    A --> F["Rule-Based Knowledge Base (Regex Matcher)"]
    E --> G["Combined Diagnostic Studio UI (Root Cause + Fix Remediation)"]
    F --> G

Loading

---

📊 Model Evaluation & Benchmarks

The deep learning models were trained and evaluated on 8,400 synthetic & real error tracebacks.

Model Architecture	Accuracy	Precision	Recall	F1-Score	ROC-AUC
Bi-LSTM (Recommended)	100.00%	1.0000	1.0000	1.0000	1.0000
LSTM	100.00%	1.0000	1.0000	1.0000	1.0000
CNN 1D	100.00%	1.0000	1.0000	1.0000	1.0000

---

📁 Repository Structure

error_classifier_progress/
├── app.py                # Streamlit Web Application (Modern Glassmorphic UI)
├── common.py             # Shared NLP Preprocessing & Vector Matrix Encoder
├── dataset.py            # Synthetic & Real Error Dataset Generator (7 Classes)
├── knowledge_base.py     # Regex Pattern Knowledge Base for Causes & Remediation
├── predict.py            # CLI Inference & ErrorAnalyzer class
├── test_analyzer.py      # Automated Test Suite (Uncertainty, Thresholds, KB Fallback)
├── train.py              # Training script for CNN, LSTM & Bi-LSTM models
├── requirements.txt      # Python Dependencies
├── README.md             # Project Documentation
├── data/
│   └── errors.csv        # Compiled Dataset
└── artifacts/            # Trained Models & Diagnostic Plots
    ├── bilstm.keras      # Serialized Bi-LSTM Neural Model
    ├── lstm.keras        # Serialized LSTM Neural Model
    ├── cnn.keras         # Serialized CNN Neural Model
    ├── vocab.json        # Vocabulary Index Mapping
    ├── labels.json       # Class Label Index
    ├── phrases.model     # Gensim Frozen Phrases Bigram Model
    ├── word2vec.model    # Word2Vec 100D Vector Model
    ├── results.json      # Benchmark Metrics Output
    ├── confusion_matrix.png
    ├── training_curves.png
    ├── model_comparison.png
    └── roc_curves.png

---

🚀 How to Run the Project

1. Install Dependencies

pip install -r requirements.txt

2. Generate Dataset (Optional)

python dataset.py

3. Train Neural Models (Optional)

python train.py --epochs 20 --per-class 1200

4. Run Test Suite

python test_analyzer.py

Validates:

• All 7 error categories on real and synthetic tracebacks.
• Confidence thresholding & "Unable to Predict" suppression.
• Input validation (empty, whitespace, non-meaningful inputs).
• Knowledge base direct matches and "No specific fix found" fallback.
• Support across all three architectures: CNN, LSTM, and Bi-LSTM.

5. CLI Inference

python predict.py "psycopg2.OperationalError: could not connect to server: Connection refused"

6. Launch Web Application

python -m streamlit run app.py

---

📜 License & Usage

Developed for Academic Capstone & Project Evaluation. Free to use and modify for research and educational purposes.