Build a decision model that flags potential fraud during an online product-booking process, based on an automatically calculated Fraud Rating Score. A booking is allowed to proceed automatically only if the score stays below 200. The score accumulates as follows:
- Booked product is a post-paid hotel: +5
- Booked product is an internal flight: +100
- Booked product is an international flight: +25
- Booked product is a car: +10
- Booked product is a pre-paid hotel: +5
- Customer has no previous orders: +100
- Customer has 1–10 previous orders: + (100 − number of orders × 10)
- Customer has previous disputes: +190
Naturally, a real fraud model would layer in many more factors — this is meant as a minimal starting point for comparing how different tools express the scoring logic.
Send your solutions to DecisionManagementCommunity@gmail.com, or open a pull request to add yours here.