Research on the Forecast of Emergency Supplies for Major Public Health Emergencies - An Empirical Study Based on the Distribution of Donated Facial Masks by the Wuhan COVID-19 Epidemic Prevention and Control Headquarters

Abstract

An adequate provision of medical supplies is critical in the battle against pandemics, such as the ongoing one against COVID-19. First, this paper proposes a generalized analysis based on the fluctuation period of emergency material demand. Second, the nonlinear problem in the low-dimensional space is transformed into a linear problem in the high-dimensional feature space by using the support vector machine method, constructing a combined forecasting model of time series and support vector machines. Lastly, the daily demand of specific protective masks donated by the Wuhan COVID-19 Epidemic Prevention and Control Headquarters in the period from February 1 to March 16, 2020 is predicted through the use of data from the Wuhan Red Cross. Compared with traditional linear time series forecasting models, the proposed forecasting model sees its accuracy increased by 37.55%, with the relative errors of mean square error, average absolute error, and average absolute error percentage being respectively reduced by 37.57%, 60.88%, and 37.86%. It transpires that the ARIMA–SVM combined model is able to make full use of the potential information implied in the original data. The decision-making process provides a reference point for the forecast of the demand of medical emergency materials in future major public health emergencies.