The Rearch And Application Of Data Mining Techniques On Medical Insurance

As in China the coverage of medical insurance is extending, the number of the people participating in the medical insurance is increasing, and the medical insurance business is more miscellaneous. In order to improve the efficiency of the medical insurance work, the medical insurance information system is widely used. A lot of data information is stored in medical insurance information system, but has not been analyzed and utilized fully. In order to discover the potential knowledge and useful information in the medical insurance data, the data mining techniques can be applied. Using data mining techniques to find the valuable information in the data is helpful for the medical insurance work implementation and hospital management.Many researchers have studied how to apply data mining techniques to solve the problems in the medical area, such as medical cost prediction, medical insurance fraud detection and medical insurance fund management. In this paper, we use data mining techniques to analyze the medical insurance data, mainly discuss two problems: frequent referral sequence mining and critical illness insurance influence factors analysis, as follows:1) Mining the frequent referral sequences, analyzing the patients’referral behavior pattern among different hosptials in a finer granularity. First the referral sequences set is extracted, then the Apriori-like algorithm is applied to discover the frequent referral sequences. For the shortcomings of the Apriori-like algorithm, this paper proposes two measures for improvement. The frequent referral sequence mining experiment shows that the the improved algorithm performs more efficiently. The valuable information is extracted according to the analysis of the frequent referral sequences mining results, it’s helpful for the management of the hosptial.2) Applying C4.5decision tree classification algorithm to analyze the influence factors of the critical illness insurance, studying the characteristics of the people with critical illness. First, the data is preprocessed according to its characteristics and problems, including data extraction, data cleaning, data discretization and class balancing, then use the preprocessed data to build a decision tree. The model generated by the experiment performs well in classification. Finally, the importance of the influence factor is analyzed and the rule is extracted according to the information provided by the decision tree.