A comprehensive fraud detection system with real-time monitoring, machine learning models, multi-agent intelligence network, and advanced AI capabilities powered by OpenAI.
- Real-time Fraud Detection: Monitor transactions in real-time with ML-powered risk scoring
- Multi-Model Ensemble: Logistic Regression, Random Forest, and Isolation Forest
- Analytics Dashboard: Comprehensive charts and performance metrics
- Analyst Review System: Manual review and feedback collection
- Model Management: Real-time model performance monitoring
- OpenAI Playground Integration: 5 AI-powered features with latest GPT models
- Multi-LLM Chatbot: Support for Ollama, OpenAI, HuggingFace, and rule-based responses
- Code Generation: Generate fraud detection code with AI
- Data Analysis: AI-powered transaction pattern analysis
- Report Generation: Automated fraud analysis reports
- Model Explanation: Explain why transactions are flagged
- Fraud Intelligence Network: Connect with other agents and systems
- Multi-Agent Communication: Real-time intelligence sharing
- Indonesian Banking Integration: BI-FAST and local bank consortium support
- Global Fraud Networks: SWIFT, Visa, Mastercard integration ready
- Clone the repository
git clone https://github.com/ghifiardi/fraud_modelling_dashboard.git
cd fraud_modelling_dashboard- Install dependencies
pip install -r requirements.txt
3. **Run the dashboard**
```bash
python3 -m streamlit run src/dashboard.py --server.port 8501
- Open your browser
Navigate to
http://localhost:8501
This dashboard is ready for deployment on Streamlit Cloud!
- Fork this repository to your GitHub account
- Go to Streamlit Cloud
- Connect your GitHub account
- Deploy the app:
- Repository:
your-username/fraud_modelling_dashboard - Main file path:
streamlit_app.py - Python version: 3.9+
- Repository:
- Live transaction monitoring
- Key performance metrics
- Real-time charts and visualizations
- Risk distribution analysis
- Individual transaction analysis
- Risk scoring and recommendations
- Recent transaction history
- Custom transaction testing
- Model performance metrics
- Transaction patterns
- Risk distribution analysis
- Customer behavior insights
- Model status and health
- Performance monitoring
- Configuration settings
- Model comparison
- Real-time alerts
- System logs
- Risk notifications
- Alert history
- Manual transaction review
- Feedback collection
- Review history
- Label management
- Multi-agent communication
- Real-time intelligence sharing
- Network configuration
- Agent status monitoring
- Code Generation: Generate fraud detection code
- Data Analysis: AI-powered pattern analysis
- Report Generation: Automated reports
- Model Explanation: Explain predictions
- Custom Prompts: Interactive AI assistance
The dashboard includes an intelligent chatbot that supports multiple LLM providers:
- Ollama (Local): For privacy-focused deployments
- OpenAI: For advanced reasoning capabilities
- HuggingFace: For open-source model access
- Rule-based: Fallback responses for reliability
| Model | Description | Best For |
|---|---|---|
| gpt-4o | Latest and most capable | Complex tasks, best quality |
| gpt-4o-mini | Fast and efficient | Good balance of speed/capability |
| gpt-4.1-mini | New GPT-4.1 variant | Optimized for efficiency |
| gpt-4.1-nano | Smallest GPT-4.1 model | Fastest, most cost-effective |
| gpt-3.5-turbo | Reliable and cost-effective | Most tasks, good value |
# Optional: For enhanced chatbot functionality
OPENAI_API_KEY=your_openai_key
HUGGINGFACE_API_KEY=your_huggingface_key- Models are automatically loaded from
models/bank_fraud_detector.pkl - Risk thresholds are configurable in the dashboard
- Real-time settings can be adjusted in the sidebar
- API Key: Enter your OpenAI API key in the Playground tab
- Model Selection: Choose from 5 different models
- Temperature Control: Adjust creativity (0.0-2.0)
- Token Limits: Control response length (100-4000 tokens)
- Real-time Processing: Sub-second transaction analysis
- High Accuracy: Multi-model ensemble approach
- Scalable: Designed for production banking environments
- Low False Positives: Optimized risk thresholds
- AI Integration: Seamless OpenAI API integration
- Frontend: Streamlit
- Backend: Python, FastAPI
- ML Models: Scikit-learn, XGBoost, LightGBM
- Visualization: Plotly, Matplotlib
- LLM Integration: OpenAI, HuggingFace, Ollama
- Data Processing: Pandas, NumPy
- AI Services: OpenAI GPT-4.1, GPT-4o models
- Jakarta Bank Consortium Agent: BI-FAST fraud patterns
- Singapore Regional Agent: ASEAN fraud trends
- Global AML Network Agent: International money laundering
- Real-time fraud pattern sharing
- Cross-border threat intelligence
- Automated alert distribution
- Network health monitoring
This project is licensed under the MIT License - see the LICENSE file for details.
- Fork the repository
- Create a feature branch
- Make your changes
- Add tests if applicable
- Submit a pull request
For questions or support, please open an issue on GitHub or contact the development team.
Built with ❤️ for secure financial transactions