Using Data Mining Methods to Detect Medical Fraud

Abstract

Medical fraudulent activities have made medical insurance expenditures rise year by year. This not only increases the burden on the medical and financial system, but also makes it difficult for many people in need to obtain these resources. Therefore, how to solve this problem has become one of critical issues. Therefore, this study aims to establish a predictive model of medical insurance fraud through data mining methods, and attempts to discover important factors affecting fraud. In this work, we will use Decision Tree (DT), Support Vector Machines (SVM), and Back Propagation Neural Networks (BPN) to establish classification models. A comparison of these three methods will be done. And, we will use decision trees to extract important factors that could provide important information for effectively detect medical fraud. Hopefully, we can effectively reduce the negative impact of medical insurance fraud.