Research On Medical Insurance Fraud Detection Based On Outlier Detection

With the popularization of medical insurance and the increasing coverage of the coverage,the number of insured persons in our country is constantly increasing,and the medical insurance business is also becoming more and more complicated.However,while the rapid development of China's medical insurance industry has made great strides,there are also various problems of medical insurance and fraud.These unlawful acts have caused great harm to the operation of China's medical insurance fund.At present,the detection of medical insurance fraud mainly rely on manual review rules and manual review,in the face of increasing medical insurance business volume,manual review alone has been clearly unable to meet the demand,then you need to use information technology to assist the audit staff to complete checking work.With the expansion of medical insurance coverage,medical insurance information system has been widely applied and developed,and accumulated a large amount of data.This provides the necessary conditions for the application of outlier detection technology in medical insurance data analysis.Outlier detection can reveal potentially meaningful information in health insurance data to aid decision making.There have been many application cases in the fraud detection of financial institutions such as banks and stock exchanges,and have good applicability.Therefore,this paper hopes to apply the outlier detection technology to the medical insurance fraud detection in our country,find the suspicious data among them,and assist the staffs decision-making.This article is mainly to medical insurance fraud detection of abnormal costs and medication problems as an entry point.Through the study of the original medical insurance data,analyze its characteristics,data pretreatment.After that,we analyzed the data characteristics based on abnormal cost and medication abnormalities,and proposed clustering-based ODC outlier detection algorithm and pruning-based OAP outlier detection algorithm respectively.After experimental comparison,they are found to be better than the original algorithm,and in the actual use of medical insurance data set has good performance.