Medical Insurance Time Series Data Statistical Analysis

In recent years,medical insurance has been the focus of attention of the state and people.Medical treatment is difficult,expensive and frequent medical disputes that have made medical reform urgent.Health care reform is a difficult problem for the whole world.It involves many aspects,especially medical management and cost control is the focus and difficulty to be addressed.Scientific and reasonable cost control can make the health insurance fund play a more full role,provide more standardized medical services for patients,and ultimately ensure the balance of revenue and expenditure of regional and national health insurance funds,thus ensuring the prosperity and stability of society.So farmany methods of c.ost settlement have been put forward internationally,For example,according to service items,according to diseases and so on,but they are not suitable for China,a country in the early stage of medical reform.However,Jilin Province is different from other provinces and regions,which forces our province to propose a cost settlement method suitable for our provinceIn the analysis of cost,we find that the ratio of person-time to number of people and the average cost are usually close to a constant,and the number of people is more stable than the cost,so it is particularly important to predict the number of outpatients.By comparing the three models of ARIMA model,Prophet model and GRU model,we find that the traditional ARIMA model can not fully consider the internal characteristics of medical insurance data,and the confidence interval of prediction is relatively large,which leads to the inaccurate prediction of small and medium-sized hospitals sometimes.However,Prophet model can fully take into account the trend,seasonality and holidays in the data,so it is more suitable for the analysis of health insurance data than ARIMA model.Finally,the GRU model is introduced.In recent years,with the deepening of in-depth learning.this model has been widely used in various fields.This model has the function of long-term memory and can achieve good predictive effect through internal learning mechanism.However,due to the limitation of the amount of medical insurance data at present,the effect of GRU model prediction is not ideal.We believe that with the accumulation of time and the increase of the amount of data,we can achieve better prediction results.In the end,this paper analyzed the forecasting results of these models,and concluded that Prophet model had better forecasting effect on outpatient visits in our province.