Document Type : Original Research Paper
Authors
1 Department of Information Technology Management, Science and Research Branch, Islamic Azad University, Tehran, Iran
2 Department of Industrial Management, Karaj Branch, Islamic Azad University, Karaj, Iran
3 Department of Information Technology Engineering, Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran
4 Department of Industrial Management, Faculty of Management and Economics, Tarbiat Modares University, Tehran, Iran
Abstract
BACKGROUND AND OBJECTIVES: It is very difficult and maybe impossible to identify suspicious damage claims in agricultural insurance using traditional methods and using the opinions of experts among a multitude of claims. In the current research, a model for discovering suspicious damage claims in agricultural insurance using data mining techniques has been presented to help the agricultural insurance fund in identifying such claims.
METHODS: The research method in the present research is applied in terms of intention and descriptive-post-event in terms of quiddity. One of the applications of data mining is anomaly detection. In the present study, a technique for detecting anomalies in the data using ensemble machine learning models is carride out. To enforcement this method, real data on compensation paid for wheat insurance (irrigated and rainfed) for one year in Khuzestan province was used. Because of differences in the process of determining damages of irrigated and rainfed wheat insurance policies, their anomalies were analyzed separately and a number of suspicious claims were acquired for each.
FINDINGS: The analysis of the results showed 5 types of suspicious behavior in claiming damages. The ratio of suspicious claims to the total (percentage of anomalies) was estimated using the histogram of anomalous scores and the opinion of insurance fund experts about 1.5%. Suspicious and unusual cases were examined by experts and the final accuracy of the model in correctly identifying suspicious cases was 72% for irrigated wheat insurance and 68% for dryland wheat insurance.
CONCLUSION: Based on the obtianed results, the presented model can be used to detecte suspicious claims in wet and dry wheat insurance policies. Since most of the unusual cases are caused by not providing sufficient documentation, it can be due to the presentation of forging insurance policies or the existence of collusion between the insured, the agent or the assessor. Therefore, more care should be taken in the payment process. The present study was conducted on the product and can be used for other crops as well.
Keywords
- Data mining
- Unsupervised machine learning
- Anomaly detection
- Ensemble learning
- Agricultural insurance
Main Subjects
Letters to Editor
Send comment about this article