The Application Of Combining Decision Tree With Logistic Regression To Analyze The Effect Of Practicing The New Type Rural Cooperative Medical System(CMS)

Objective:This article is for the purpose of analyzing data of tracing investigation after implementing the new rural cooperatives medical service in the experiment site, which using the data mining method which unifies the decision tree and the Logistic regression analysis, to discusses the practical application feasibility, the concrete implementation method and the reality application value of decision tree analysis method ,which use in implementation effect correlation factor research of the new rural cooperatives medical service system.Methods:Investigations were completed in three counties of Jiangxi Province to investigate the implementation effect of the new type rural CMS by the method which unifies the typical survey and stratified cluster sampling in the year of 2006; Through establish the goal database after the data sorting , with the aid of Enterprise Miner platform of SAS software, according to the SEMMA data mining theory, to establish the decision tree model to analyze the local farmer's will to participate the new type rural CMS, the optimized Logistic regression analysis model , the decision tree model to analyze the influence factor of choosing medical institution of local farmer and the multivariable Logistic regression analysis model, and carry on the comparative analysis between the tree of will and the fitting of regression model before and after optimization, combine the effect good tree model with the optimized regression model to analyze the farmer's will. All of which can be used to discuss the influence factor of choosing medical institution of local farmer scientifically.Results:CART tree model is the best tree model to analyze the local farmer's will to participate the new type rural CMS, its analysis result prompts the influence factors are: the understanding to the fair of reimbursing, the form of expense burden, the attitude and educational level of personnel who manage reimbursing, whether knows others reimburse or not, whether knows publication or not, the appraisal to medical quality, a distance to the nearest medical institution. But the optimized Logistic regression model which combine with the tree model shows the influence factors are: the understanding to the fair of reimbursing, the form of expense burden, the attitude and educational level of personnel who manage reimbursing. The variables selected by the best tree model include which selected by the optimized Logistic regression model. Through the analysis of the tree chart and its classified rule, we conclude that the understanding of fairness in the expense burden has the different influence to the will of join the new type rural CMS in crowd with different form of expense burden.CART tree model is the best tree model to analyze choosing medical institution of local farmer,both CART tree model and one-way regression analysis select twelve variables into the model; multivariable Logistic regression which Unifies these variables shows that the form of expense burden and gender are rejected in the final Logistic equations; Finally we combine the variable important value which is provided by tree analysis with parameter estimation value provided by Logistic regression , then the finding points out that the type of disease is primary factor to influence farmer to visit doctor, next is educational level and occupation. The compensation proportion gradient strategy of each level medical institution don't adjust farmer's choice of medical institution reasonably, while the educational level of farmer and grasping certain medicine knowledge are forward promotion factors, and at the same time it is necessary to the development and stabilization of the new type rural CMS that we should ensure the farmer who go out for working can receive treatment, and it is worth studying.Conclusions:This article combine decision tree model with Logistic regression to carry on analyzing the will of farmer to participate the new type rural CMS and the influence factor to their choice of medical institution, simultaneously we make detailed description to this data analysis process. Through comparing these two analysis method, we discover that both have its superiority respectively in the data analysis process, the best method is composing merit organic union in the actual data analysis process, displaying two methods'superiority to the maximum extent, but should not confine to discuss which is fit or unfit.