The Study And Application Of Forecasting Model On The Single Disease Medical Expenses Based On Elman And SVM-With The Sample Of Colorectal Cancer

ObjectivesBy collecting, analyzing the medical expenses data from the hospital management information system of a hospital, single disease medical cost data warehouse was constructed, then single disease medical expenses data was discussed and processed, prediction model of a single medical cost effective disease was eventually established. It is a preparation and forward-looking work for the reference system of a single disease costs model.Methods:The patients were selected from HIS system in the medical record and data integration, access to, the application of principal component analysis of the data had been cleaned, a large variable factors were selected to analyze. Processed data were statistically analysed and modelling by SPSS18.0Results:①Three common factor were concluded by factor analysis, namely:treatment factor, payment factor, demographic factor;②More than 69.75% of the original data information can be described by three common factorsConclusion:According to factor in the cost model, using SVM to classify the training, it was found that it had the highest accuracy of the classification according to the surgical category (yes or no), which indicates that the cost of a single disease colorectal cancer sample data are classified in accordance with the procedure most favorable category subsequent prediction; Upon the SVM, building Elman neural network forecasting model showed the highest prediction accuracy of SVM-Elman model portfolio, up 98.2%. with a separate SVM, BP neural network, it was found that combined model was most suitable for the colorectal cancer single disease cost forecast.