Research And Implementation Of Cost Forecasting Model Based On DRGs Grouping

In recent years,China has faced many problems,such as increasing total medical and health care costs,irregular disease diagnosis and treatment process,difficult and expensive medical treatment for the common people,and serious imbalance between income and expenditure of medical insurance fund.In view of the remarkable achievements of disease diagnosis-related clusters(DRGs)in the field of medical cost control at home and abroad,this paper combined machine learning algorithm to carry out the research on the cost prediction model based on DRGs grouping,so as to provide a new reference for the control of unreasonable growth of medical costs and a new idea for disease cost prediction methods.The main research work of this paper is as follows:First of all,this paper takes the data of the first page of the medical record of a third class A hospital mainly diagnosed as CHD as the research object,and carries out a series of pre-processing operations such as data integration,protocol,cleaning and transformation,in order to improve the data quality and ensure the rationality of the subsequent DRGs grouping results and the scientific nature of the cost prediction model.Second,the grouping principle and grouping logic of CHS-DRGs are deeply studied in this paper.Mono factor analysis and multiple linear regression were used to analyze the main factors affecting the hospitalization cost of CHD,which were used as the grouping node of DRGs.The CART,CHAID and E-CHAID decision tree algorithms are studied deeply,and the DRGs grouping is carried out through these three different decision tree algorithms,and the optimal grouping model is selected.The results showed that E-CHAID-based DRGs had the best grouping effect,with higher intra-group homogeneity and greater inter-group variability.Then,this paper proposes a GWO-SVR cost forecasting model based on DRGs grouping.By eliminating the data of patients outside the line,the weight of diseases grouped by DRGs was used as the input feature to predict the cost,which enhanced the reliability of the prediction results of the model and played a role in reasonable cost control to a certain extent.The prediction accuracy of SVR algorithm is guaranteed by selecting the appropriate kernel function.The gray Wolf algorithm was used to optimize the parameters of SVR to solve the defects of high operation and high time consumption of SVR algorithm in large-scale samples,reduce the time of parameter optimization and further improve the prediction accuracy.The superiority of the proposed GWO-SVR algorithm is verified through experimental comparison with the SVR models of other optimization algorithms.The results show that the GWO-SVR cost prediction model based on DRGS-grouping can well predict the hospitalization costs of coronary heart disease.Finally,from two aspects of system design and system implementation,the Spring MVC framework and My SQL database are used to realize the cost prediction system.