Explainable Fraud Detection in Open Banking: 98.7% Precision

This is a portfolio project from ZenDev Co — an Open Banking Payments Intelligence Platform built to demonstrate how explainable fraud detection systems can be constructed.

We developed a custom anomaly detection system combining TensorFlow for pattern recognition with FastAPI for real-time processing. The system uses unsupervised learning models to detect unusual transaction patterns across 12 UK banks, with a custom dashboard that provides clear explanations for every alert — enabling compliance teams to understand why a transaction was flagged faster.

The final system achieved 98.7% precision while processing over £2M monthly across the 12 banks. It is designed to reduce false positives and accelerate compliance teams in understanding why transactions were flagged.

This project proves that explainable fraud detection is achievable without sacrificing accuracy.