Research On Anomaly Detection For Medical Insurance Record Based On Improved Fuzzy Clustering Algorithm

There has been an increasingly urgent problem in the medical industry,that is,the large number of insured and the high amount of medical insurance funds are facing extremely serious risk of violation.Every year,the loss of medical insurance fund caused by fraud has exceeded 10 billion yuan.The national audit office once publicly stated in the Standing Committee of the National People's Congress that "the problem of medical insurance fraud is a very serious challenge".Due to the overall limited medical insurance funds,those who violate the rules directly occupy the rights and interests of the real patients,and may even cause painful medical incidents.Existing fraud detection work of medical insurance is often carried out by special auditors in the way of manual verification and screening,which is not only inefficient,but also inevitably has the situation of missing and wrong inspection.Although information technology can not accurately locate the violation records,it can assist the staff in the first step of large-scale screening,lock the abnormal records through outlier detection,narrow the scope of subsequent secondary screening,and greatly reduce the amount of manual work.Clustering based outlier detection algorithm has the characteristics of low time complexity and no need for high-quality training set.However,the effect of this algorithm largely depends on the accuracy of the description of the objective things,and it is easy to fall into the local optimum when the amount of different types of data is very different,which has a significant impact on the accuracy of clustering results and detection.Based on the detailed analysis of the traditional clustering algorithm,this paper proposes an interval image fuzzy clustering algorithm based on data volume,and applies it to the work of medical insurance anomaly detection.The main research contents are as follows:(1)Research on the actual medical insurance audit methods,and analyze the pain points of the existing research.Generally speaking,the medical insurance audit mode that relies on artificial experience for screening takes the clinical path of the disease as a reference,that is,the same type of disease has similar diagnosis and treatment measures and costs,while those records that differ greatly from most clinical paths will be regarded as abnormal records suspected of fraud.This principle is similar to the idea of outlier detection in machine learning.Based on the research results,this paper analyzes the existing outlier based health insurance anomaly detection research,selects clustering algorithm as the research method,and studies how to information,improve the description of reimbursement records,so as to improve the clustering effect.(2)According to the unbalanced characteristics of medical insurance data set,the clustering algorithm is improved.When preprocessing and analyzing the medical insurance data set,it is found that the amount of abnormal records is very small compared with the normal records,and this imbalance attribute will produce less data sets in the traditional clustering algorithm,which will be misclassified to a large data set to a large extent,thus affecting the clustering results.Aiming at the problem of poor clustering effect of existing clustering criteria on unbalanced data samples,this paper introduces the concept of data volume and proposes a new clustering criterion(imbalance aimed index,IAI),which can produce better clustering effect in balanced and multiple types of unbalanced data.(3)Providing an anomaly detection model of medical insurance based on improved fuzzy clustering algorithm proposed.In the medical field,the traditional hard division limits the possibility of using data to describe the actual situation of reimbursement records in a more detailed way,and the introduction of fuzzy theory solves the problem of information distortion.In this paper,by introducing the more advanced interval image fuzzy theory,an improved fuzzy clustering algorithm based on c-harmonic clustering algorithm is proposed.It not only improves the local optimal problem of traditional fuzzy clustering in outlier detection,but also considers the uncertainty of medical records in classification,so it has a wider application range and stronger robustness.By using the real medical insurance reimbursement data set to verify the proposed detection algorithm,the integration operation and clustering results of medical insurance records in the interval image fuzzy environment are given.The experimental results show that the algorithm can effectively judge the abnormal medical records and has a good detection rate.