The definitive ecosystem for high-speed transaction monitoring and behavioral risk analysis.
In a world where digital fraud evolves every second, static rules aren't enough. PayGuard AI bridges the gap between Complex Neural Mathematics and Actionable Business Logic.
- 🚀 Performance: Sub-15ms inference latency.
- 🛡️ Accuracy: Ensemble modeling catching 99.8% of fraudulent signatures.
- 🇮🇳 Localized: Custom support for Indian regional transaction patterns and ₹ (INR) formatting.
- 🧠 Transparency: Explainable AI (XAI) tells you the "Why" behind every flag.
| Engine | Best For | Logic Type |
|---|---|---|
| XGBoost Pro | Maximum Accuracy | Gradient Boosted Trees |
| Neural Network | Rare/Smart Fraud | Deep Learning (Focal Loss) |
| Ensemble Forest | Complex Strategies | Random Forest (Bagging) |
| Linear Logic | High-Speed Rules | Logistic Regression |
PayGuard-AI/
├── 📂 Backend/ # Python Flask Intelligence Layer
│ ├── 📂 api/ # API Blueprints (Predict, Upload, Dash)
│ ├── 📂 data/ # Synthetic Dataset Generators
│ ├── 📂 models/ # Model Training & Loss Logic
│ ├── 📂 saved_models/ # Compiled Model Weights & Scalers
│ └── 📂 utils/ # Preprocessing & Metric Logic
├── 📂 Frontend/ # React High-Performance UI
│ ├── 📂 src/
│ │ ├── 📂 components/ # Liquid Glass UI Kit
│ │ ├── 📂 hooks/ # Custom React State Logic
│ │ ├── 📂 lib/ # API/Fetch abstraction
│ │ └── 📂 pages/ # High-fidelity View Layers
├── 📄 .gitignore # ML-Optimized Git Control
├── 📄 README.md # Ultimate Project Documentation
└── 📄 start.bat # One-click system launcher
POST /api/predict
{
"Time": 3600,
"Amount": 5000.00,
"V1": 1.5,
"model_type": "xgboost"
}POST /api/upload
- Format: Multipart/Form-Data (CSV)
- Engine: Handles thousands of rows via chunked processing.
cd Backend
pip install -r requirements.txt
python models/train_models.py # Initialize weights & scalers
python app.py # Launch API on port 8000cd Frontend
npm install
npm run dev # Launch UI- AI-Powered Auto-Blocking: Connect to payment gateways for instant refund triggers.
- Global Geo-Fencing: Map-based visualization of regional threats.
- Model Retraining UI: Feed user-reported "False Alarms" back into the model for live learning.
PayGuard AI is an advanced simulation tool for fraud intelligence. Always verify transactions through official banking protocols before taking legal action. The data used is synthetic and designed for behavioral analysis demonstration.
Developed with Intelligence by Mourique Naskar