Case study: Health Claim Fraud Detection for a Leading Insurance firm
About the Client
The client is a leading insurance provider dealing with multiple insurance products for individuals and businesses.
The client was dealing with higher triage overheads due to low accuracy of existing fraud detection system for health claims. The system lacked ability to use relationships between entities and advanced AI algorithms.
BRIDGEi2i’s fraud management solution helped reduce false positive by ~30% using multi-layered strategy to detect fraudulent claims. Investigators were also assisted with reason for alert as well as potential investigation path thereby reducing turnaround time for investigations
BRIDGEi2i leveraged its data engineering expertise to extract data using pre-defined ETLs, transform data using our feature generator and load information onto our platform.
Data used included not only structured data on customer demographics, personal, policy, claim and medical information, but also unstructured data around relationship between various parties associated with the claim as well as description of claim including claimant notes.
We developed multi-layer strategy to identify fraudulent claim activity using:
- un-supervised model for localized and globalized anomaly detection
- supervised model for identification of historic frauds
- graph networks for identifying collusions and investigation path
by generating features from text, relationships and structured data.