Prediction Of Total Health Expenditure In China Based On PCA-BP Neural Network Model

The total cost of health refers to the total value of the health resources spent in a certain region to provide health services for the residents in a specific period of time,and is also the total amount of resources spent by a country or region to implement measures such as treating diseases and promoting health.The total health expenditure reflects the macro situation of the development of the medical and health industry in a country or region and the degree to which the government,society and individuals attach importance to the medical and health industry.It is an important indicator used to evaluate the health fund preparation,health fund input,health resource allocation and other aspects,and is regarded as the GDP of the medical and health field.Therefore,it is necessary to forecast the total health expenditure in China.This paper first analyzes the situation of financing and flow direction of total health expenditure in China from three aspects of financing level,financing structure and institutional flow direction.Based on the theory of supply and demand of health services and the theory of health demand,combined with China’s macro health production function,The influencing factors of the total health expenditure were divided into five categories: economic income factor,social population factor,health service supply factor,health service demand factor and education factor,and 15 indicators were selected to construct the influencing factors system of the total health expenditure in China.Secondly,the data of China’s total health expenditure and influencing factors from 1980 to 2019 were selected,dimensionality was reduced through principal component analysis,and the normalized principal component score was input into BP neural network to construct a scientific and reasonable PCA-BP neural network model.Compared with BP neural network model without principal component analysis,traditional ARIMA model and polynomial trend curve prediction model,it is found that the PCA-BP neural network model has the least prediction error.The mean absolute error,mean relative error,mean square error and root mean square error are 7.10,0.00011,147.39 and 12.14,respectively,which are suitable for predicting the total health expenditure in China.Finally,the PCA-BP neural network model was used to forecast the total health expenditure of China in the next three periods,and the predicted results were 71,892,77,764 and 82,578 billion yuan respectively.Then from reducing the proportion of personal health expenditure,giving full play to the basic role of primary medical and health service institutions,and combining on-site diagnosis and online diagnosis and treatment,this paper puts forward the countermeasures and suggestions of using the total health expenditure in practice.The first innovation of this paper is to construct the index system of influencing factors of the total health expenditure in accordance with the actual conditions of our country by combining with China’s macro health production function.The second is the combination of principal component analysis in multivariate statistical analysis and BP neural network model based on machine learning,which can not only eliminate the correlation of input variables,but also simplify the network structure and reduce the prediction error on the basis of retaining the original information.