| title: Driver and Fraud Risk Hub | |
| emoji: 🛡️ | |
| colorFrom: purple | |
| colorTo: red | |
| sdk: docker | |
| app_port: 8501 | |
| tags: | |
| - streamlit | |
| - classification | |
| - fraud-detection | |
| - risk-modeling | |
| pinned: false | |
| short_description: Driver claim and transaction fraud risk screening | |
| # Driver & Fraud Risk Hub | |
| Two interactive risk-modeling projects: | |
| - Porto Seguro driver claim probability screening; | |
| - PaySim mobile-money transaction fraud-risk screening. | |
| The fraud model was retrained without account names or balance fields and evaluated on 64,554 unseen transactions, reaching **0.971 ROC-AUC** and **87.8% fraud recall**. Outputs prioritize cases for human review; they are not accusations, insurance quotes or automated eligibility decisions. | |