Analysis of TVS warranty claims and prediction of fraud

Abstract

Our objective of this study is to analyze patterns in the data provided by TVS dealers regarding the claims processed by them. We intend to find associative patterns in the dataset which will enable TVS to detect possible sources of fraud, predict the problems which are occurring repeatedly and are connected, and also forecast demand for specific spare parts which are large in number. The general scope of the project includes developing models for detection of fraud in warranty claims of vehicles or other consumer durables like Refrigerators, Air Conditioners, Laptops, etc. On an average fraudulent warranty claims results in 3-15% of the company’s total warranty costs. Hence, a robust fraud detection model can significantly reduce a company’s warranty expenses. Similar models may also be applied to insurance industry to detect fraudulent insurance claims.