【作者】 王凯；

【导师】 魏善沛；

【作者基本信息】 中南林业科技大学 ， 计算机应用技术， 2012， 硕士

【摘要】 社会医疗保险是社会医疗保障的基础,坚持“低水平,广覆盖”的原则,强调“公平优先、兼顾效率”,承担基本医疗的保障职责。我国对社会医疗保险的投入一直在不断的增长,社会医疗保险领域的信息化建设正在迅速发展。社保医疗信息管理系统的建设,为医疗基金的管理带来了方便,但是随着参保人数以及待遇项目不断增多,目前,不同地区的医保信息系统里已经积累了大量历史数据,如何对这些数据做一些有效的分析,已经成为现阶段数据管理的一个难点。随着数据挖掘技术的不断成熟与发展,其对数据的处理功能已经被人们接受并且广‘泛应用到有数据分析需求的行业之中。在医保信息管理方面引入数据挖掘技术,对大量沉积的历史数据进行处理,得出一些没有发现的信息和规则,辅助医保基金风险防控和安全运营。本文首先总结了目前在社会医疗保险管理上遇到的问题。介绍了常用的数据挖掘方法,对聚类分析进行深入的探讨。研究了EM算法,并在选取初始化参数和最大化步骤上对EM算法进行改进。在SQL Server2008数据挖掘平台基础上,实现改进EM算法,并通过插件方式注册到分析服务器算法库中,通过与EM算法的比较,得出改进EM算法的有效性。随后,在SQL Server2008上创建面向医疗消费费用异常检测主题的数据仓库,生成多维数据集,用改进EM算法创建数据挖掘模型,在此模型上,利用predict函数取得各种医疗消费费用的预测值,并通过一系列的数据分析规则,得出医疗消费费用的异常记录。同时,标注真实的异常消费记录作为跟踪,比较改进EM算法在查找异常记录上的查准率,证实了该算法在实际应用中的有效性。本文的内容作为一种尝试性的探测,得出一种取得医疗费用异常消费记录的分析方法。本文主要研究成果包括：(1)改进EM算法,并验证改进EM算法的有效性；(2)创建面向医疗消费费用异常分析主题的数据仓库,生成多维数据集,并在此基础上完成数据分析流程,取得费用的异常消费记录,成为一种有效的分析消费费用异常的方法。更多还原

【Abstract】 Social health insurance has been assumed the protection of basic medical duties. With the continual investment of social health insurance, the management system of social health insurance is developed fast. The information management system of social security provides convenience for the management of social security. The number of insured people is increased, and different parts of the social security system have been accumulated a large amount of historical data. With the rapidly development of data mining technology, the theories are became matured, the data processing functions have been accepted and applied to data analysis. The introduction of data mining techniques in the social security management, excessive sedimentation of historical data for processing, to find out some hidden information and rules in large data sets, also, to prevent the risk of social security fund and control the security operations.In this thesis, the author first summarizes the problems sometimes encountered in the management of the social security medical insurance. Then, the theory of data mining methods, and cluster analysis is discussed. Later, introducing EM algorithm process, using the improved EM algorithm to select the initial parameters and maximize step simplifies the EM algorithm. SQL Server2008data mining platform on the basis of improved EM algorithm and plug-ins registered to the analysis server algorithm library, by comparing with the EM algorithm, improved the validity of the EM algorithm. Subsequently, the data warehouse for medical expenses anomaly detection topic in SQL Server2008to be created, by using a modified EM algorithm to create data mining models, this model, predict function to obtain the predictive value of various medical and consumer costs and data analysis through a series of rules, drawing the exception record of the costs of medical consumption. At the same time, the true abnormal consumption records are improved the EM algorithm and found out the exception record precision, the validity of the algorithm in practical applications is confirmed. The contents of this article obtained an analytical method to obtain the medical costs of abnormal consumption record.The main research of this thesis are as follows, firstly improved the EM algorithm, and validation to improve the effectiveness of the EM algorithm, secondly, created the data warehouse and generated the cube in the themes of detect the abnormal information in medical consumption of social security, and completed data mining on this basis analysis process, at last we have got the conclude that the result was useful. It’s an effective method of detect medical consumption abnormal information.更多还原