Predicting Sham Claims in Vehicle Insurance using Machine Learning Techniques

The insurance industry is experiencing rapid growth and as a result, faces the challenge of managing vast amounts of data. One of the foremost concerns within this sector is the issue of fraudulent claims, which refers to deceptive attempts made for personal financial gain. This deceitful practice affects both insurance companies and their honest policyholders due to the prevalence of such claims. Traditionally, domain expertise and conventional methods have been employed to detect fraudulent activities in insurance claims. However, more recently, data mining techniques have emerged as valuable tools in the field of insurance analysis. These techniques enable the calculation of insurance premium amounts and the prediction of both fraudulent claims and accident occurrences. The premium amount is determined based on individual customers’ confidential and financial information. Implementing this research significantly reduces human effort and financial losses in the examination of automobile insurance claims. Effective classifiers play a crucial role in determining whether a claim is fraudulent or not. Predicting the occurrence of accidents is a fundamental step in identifying fraudulent claims. Decision tree classifiers, such as Simple CART, SVM, and LMT, available in the WEKA tool, are employed to analyze accidents and claims. The execution of these classifiers is compared in terms of accuracy, with Random Forest demonstrating superior accuracy in predicting accident occurrences and yielding the best results among the classifiers. Furthermore, the SVM algorithm outperforms other classifiers in predicting fraudulent claims.