Identification Of Fraud BP Neural Network-based Health Insurance

Insurance fraud cases have been increasing year by year globally, which has resulted in hundreds of billions of dollars loss each year. Insurance fraud can distort the pricing mechanism of insurance, damage insurance utmost good faith principle, threat to the security of insurance fund,impede the implementation of health insurance policies. Therefore, the anti-fraud work has become a national important issue to maintain the insurance industry to develop normally. While insurance fraud recognition research has become a hot point in the field of insurance at home and abroad. The theoretical study on insurance fraud has been more in-depth. The researchers have studied the forms and causes of fraud, anti-fraud measures systematically from views of information economics and social psychology. The empirical research has developed from statistical analysis methods to artificial intelligence recognition technology and combination of the two. By health insurance information technology and data limitations reserves, and the hinder of complex medical environment, the research on health insurance fraud identification and measurement is still relatively week in the domestic compared with the study abroad.With reference to motor vehicle insurance fraud recognition research in domestic, this paper tried to combine the statistical regression with neural network together to build a model for the recognition research of health insurance fraud arising from moral hazard. In this paper, we analyzed the issue of health insurance fraud with reference to the related research at home and abroad. We collected the data of hospital medical insurance claims as experimental samples. On the basis of theoretical analysis and according to experts’ opinion we chose several identify factors. Then the factors were refined by logistic recognition.The BP neural network model was built and trained using sample data before validity test. Based on the results the targeted anti-fraud measures and policies recommendations were proposed. The results showed that BP neural network model which was embedded with logistic regression can be used as an effective tool to identify fraudulent health insurance claims under certain conditions. This study will reveal some fundamental characteristics of health insurance fraud to a certain extent, and has practical significance for improving fraud recognition technology and anti-fraud capabilities of insurance institution.