Case Study
Risk Analytics
Credit Bureau: Develop a robust Application Fraud prediction model
Developed machine learning based fraud model, using only credit bureau header data as input features and address mining
Created national fraud host spots
Implementation of explainable machine learning output for user codes
Effective at predicting fraud
for large multinational company lenders and smaller non-banking financial companies
Credit Risk Model Approach
- Acquisition Score
- Credit risk assessment of new customers
- Acquisition Risk strategy and scorecard implementation
- Identity resolution and verification
- Behavior Score
- Ongoing – monitoring, CLI/CLD, Top-Up, Cross Sell
- Transaction control
- High risk account management
- Pre-delinquent collection
- Collection & Payment Score
- Recovery & Repossession
- Projected payments
- Collection optimization
Results
87% Accuracy in Capturing App Fraud
Other Case Studies
Customer Segmentation
M&A Evaluation
