The Application Of Decision Tree In The Information Mining Of Health Service

Objective: The main purpose of this paper is to discuss the operating flow of data mining, such as decision tree and regression, and its possibility of being applied to the information mining of health service. We use the health seeking rate to explain the need of health service, the clinic medical expense to explain the using of health service.In this way, we can mine the main reasons for people's health seeking rate and clinic medical expense to know more about people's behavior of their health seeking and to provide information for health decision. Method: According to the data of the third national health service survey, we use Enterprise Miner, a special data mining software of SAS Company, to set up the classified tree model and the Logistic regressive model of people's health seeking rate, as well as the regressive tree model and multivariate regressive mode of clinic medical expense, on the basis of the standard of SEMMA. We also compare the model fit statistics and analytical results of the relevant decision tree with regressive models, and choose a more effective model as the final model. Result: We build decision tree model for people's health seeking rate, and the classifier error rates for the training and validating data sets are 0.366 and 0.383. Meanwhile, we set up Logistic regressive model, and the classifier error rates are 0.387 and 0.385. So we choose decision tree model which is fewer in classifier error rate as the final model. Therefore, the reasons for people's health seeking rate are people in urban or rural areas, Engel index, the ponderance of illness, the health seeking distance, profession, the chronic diseases, educational degree, income, ethnic group and age, etc. The Root ASE of the training and validating data sets are 1.2237, and 1.2662 when building decision tree for people's clinic medical expense. However, the Root ASE becomes 1.2513 and 1.2516 when setting up multivariate regressive model. So the reasons for clinic medical expense are the ponderance of illness, the chronic diseases, medicare, Engel index, age, profession, income, and health seeking distance, etc. Conclusion: From the cases and applications in this paper, We have finished the process of data mining and made description and explanation of it. Comparing decision tree with regressive analysis, generally speaking, the results of the two ways of analyzing do not differ too much from each other, relatively speaking,the decision tree is better than the regression. Additionally, we can find out that the causes selected by the means of decision tree have different effects on different people.