Document Type : Original Research Paper

Authors

• Z. Seidi Aghil Abadi ¹
• S. Sehhat ²
• R. Salehi ³

¹ Department of Business Management, Department of Insurance Management, Islamic Azad University, Science and Research Unit, Tehran, Iran

² Department of Business Management, School of Management, Allameh Tabatabai University, Tehran, Iran

³ Automobile Insurance Group of Sina Insurance Company, Tehran, Iran

Abstract

The insurance industry, by its very nature, is susceptible to fraud. In car insurance, the insurer covers all damages caused to third parties by the car or car load. In recent years, due to the growth of this type of insurance, it has become necessary to identify the influencing factors on the decisions that deal with the falsity of a damage claim. One of the ways to detect and deal with this type of fraud is to check the information in the files that have claimed damages through a third party insurance policy. Data mining is a suitable method to interact with such databases and leads to the discovery of valuable knowledge from them; In this research, by examining 142 third party cases and 6 variables, it has been tried to discover fraud patterns in third party insurance. The research results show that the decision tree algorithm and neural networks have performed better than the support vector machine algorithm in identifying fraudulent, non-fraudulent, and suspicious cases.

Keywords

• Fraudulent claim
• Data mining
• Classification
• decision tree
• Support Vector Machine
• Neural Net
• Fraud