Sanntidsdeteksjon av avvik i finansielle transaksjonsstrømmer på tvers av flere banker og ERP-system

Offentlig prosjektsammendrag

This project aims to develop real-time machine learning models for detecting payment anomalies across the traffic from and to 250+ banks and 25+ ERP systems. The core R&D challenge involves developing confidence-adjusted models that predict normal payment behavior under conditions where processing volumes, timing, and system interactions vary significantly across different bank-ERP combinations. The project generates new understanding of mathematically modeling "normal" payment traffic behavior. The proposed solution systematically addresses identifying features from incoming and outgoing payment data and adapts anomaly detection algorithms for multi-system environments while maintaining low false positives. It develops interpretation frameworks that can distinguish critical failures from normal variations and creates automated response systems capable of safely executing corrective actions across diverse banking partners.