Insurance Claim Analytics
A leading Insurance company wanted to develop an analytical tool to accurately identify fraud claims for motor insurance. The manual process of fraud claim identification was increasing the chances of errors like missing or wrong calculations, providing more time to the fraudsters and related with high cost.
The objective was to develop a set of rule or indicators that can flag potential fraud claims as soon as they registered in the system. The ultimate aim was identified as development of a scoring engine using key indicator to predict/identify the suspicious claims.
The data used for the analysis was mainly claim behavior, Claimant’s behavior and profile, Demographics, Historical policy information, macro-economics and circumstantial story. Once the business objective was identified, we used statistical methods like correlation, association rules, clustering of variables etc. to identify the key drivers of fraud behavior.
Based on the distribution of fraud claims and the relationship of different key drivers with the target, logistic regression was found most suitable to serve the business objectives.
The process of identification/flagging of the suspicious claims was fully automated. Now the client can use the score given by the model to tag very risky and costly claims. The claims with higher risk score are given preference and investigated more rigorously. The model is able to predict fraud claims with >70% accuracy.