CustomerXPs: AI-based Hybrid Approach to Banking Fraud Detection and Prevention
Posted: 23 May 2018 | Author: Jayaprakash Kavala | Source: Clari5
The use of hybrid fraud detection models to ensure high fraud detection rates with low false positives is a vital aspect of enterprise fraud management. A
‘hybrid’ fraud detection model comprises 4 primary techniques, viz. pre-packaged complex scenarios, behavior profiling, link analysis, and lastly machine learning based predictive risk scoring. These techniques are used to accurately risk score every transaction and to advise the right action of whether to approve, decline or challenge the transaction in realtime with sub-second response time.
The ability to analyze massive transaction data and build fraud risk intelligence in real-time is a critical capability in this context. Querying the massive
transaction volume for behavior profiling, entity link graph building, and training predictive scoring models on the large dataset are of paramount importance in fraud detection.