Card fraud detector

The brief

A card issuer sees one fraudulent transaction in several hundred, and its rule-based alerts annoy genuine customers. Build a model that catches more fraud with fewer false alarms, and explains each alert so the fraud team can act quickly. Good looks like: metrics suited to extreme imbalance, a threshold chosen by cost, and an explanation for every alert.

What you will build

1. Explore the imbalance and the features
2. Train gradient boosting with class weights or resampling
3. Evaluate with precision, recall and PR-AUC, not accuracy
4. Choose a threshold from the cost of fraud and of false alarms
5. Explain alerts with SHAP

Free, open data to build it with

• Credit Card Fraud Detection ↗ Kaggle (ULB)
• IEEE-CIS Fraud Detection ↗ Kaggle

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.