Research On Suspected Fraud Identification Of Commercial Medical Insurance Based On Data Mining

Recently,China's medical insurance fraud problem is very serious,but the medical data of major medical institutions cannot be fully shared,so there is little research on medical insurance fraud and fraud identification.This paper uses the medical database owned by consulting companies and insurance companies to identify suspected fraudulent behaviors of policyholders and medical institutions by using data mining methods,and establishes suspected fraud identification models for policyholders and medical institutions,which can be applied to actual business.Medical insurance fraud identification,and provide reference for social medical insurance fraud identification research,to achieve the purpose of reducing medical resources waste.Policyholders who intend to commit fraud will have specific suspected frauds during the medical treatment and claims process,and thus will have unique regularity and unique behavior.This paper builds a Logit regression analysis model to find out which feature factors are highly correlated with fraud.The highly relevant feature factor is turned into the input variable of the Back-Propagation backpropagation network model.The constructor and the running result of the model are used to judge whether the policyholder has suspected fraud.The form of fraud committed by medical institutions is mainly in collusion with patients(insurers).There are many types of fraud in medical institutions.This is mainly because of the complexity of the medical process.The medical institution(doctor)can make a claim for diagnosis,fraudulent prescribing,deliberately extending the length of hospital stay,agreeing to different patients using the same insured person's information,or breaking a medical treatment process into multiple treatments to increase the amount of the claim,etc.Therefore,the fraud of medical institutions is more diversified and more varied than that of policyholders,it is also limited due to the length of this paper.This paper chooses to identify whether the medical institution is acting on the fee list to test whether the medical institution has suspected collusion with the policyholders,as an example to provide an idea for the identification research of the conspiracy fraud model of medical institutions.Therefore,in order to deal with this type of fraud,this paper establishes the K-Means clustering model to identify the fraud according to the idea that the medical insurance claim amount and expenditure meet Benford's law.Finally,the suspected fraud identification model for policyholders and medical institutions can be used to identify fraudulent behaviors after actual data verification.This paper proposesthat China should improve the medical insurance legislation system,establish and improve the key indicators of medical insurance fraud identification,and gradually share medical data,development and promotion of medical insurance fraud intelligent identification technology and strengthening cooperation mechanisms between insurance companies and medical institutions,and strengthening practical advice on medical behavior and medical expenses.