DBT Transaction Monitoring System 🛡️

Team ID / Name: TS423 - Eternal Champions

Problem Statement

Gujarat disburses welfare benefits through Direct Benefit Transfer across multiple schemes. Known failure modes include:

• Funds reaching ineligible or deceased beneficiaries
• Undrawn funds (beneficiary never notified)
• Duplicate payments under slightly different names or across schemes

Features

This system processes 50,000+ simulated DBT transactions and generates a structured audit report. It includes:

1. 4 Leakage Pattern Detectors: Deceased beneficiary, duplicate identity, undrawn funds, and cross-scheme duplication.
2. Fuzzy Name Matching: Handles Gujarati transliteration variations.
3. Risk Scoring: Assigns explainable risk scores and specific evidence for each anomaly.
4. Prioritized Queue: District Finance Officers get an interactive queue instead of a raw data dump.
5. High-Speed Processing: Powered by Polars, processing 10,000+ transactions in under a second.
6. Multi-Role Portals: Streamlit UI supporting DFOs, Scheme Verifiers, Audit Team, and State Admins.

Requirements

• Python 3.9+
• Dependencies: polars, rapidfuzz, plotly, streamlit

How to Run

pip install polars rapidfuzz plotly streamlit
streamlit run app.py