Spatio-temporal Analysis And Risk Prediction Model Of COVID-19 Transmission In China

Since the end of 2019,the novel pneumonia epidemic has spread rapidly in various countries around the world,making global public health security not optimistic.Therefore,this paper establishes a risk prediction model based on the Spatio-temporal characteristics and risk factors of COVID-19 transmission in China.Data sets from 31 provincial-level administrative regions in China(excluding Hong Kong,Macao,and Taiwan)and five different sources were collected for analysis,including confirmed COVID-19 case data,Baidu Migration Index data,and Government Response Rigor Index data,air pollutant data and meteorological data.First,spatial autocorrelation,hot spot,and Spatio-temporal scanning statistical methods were used to explore the Spatio-temporal characteristics of COVID-19.Secondly,univariate and multivariate analysis methods in statistics were used to explore the relationship between epidemic risk factors and confirmed COVID-19 cases.In this study,ARIMA(Autoregressive Integrated Moving Average Model)and LSTM(Long short-term Memory network model)were used as a regressive Integrated Moving Average model.Statistically,significant risk factors were included to build a risk prediction model for COVID-19 transmission in China to simulate,verify and predict the risk of COVID-19 occurrence in different regions.The results are as follows:(1)The regions with the highest number of confirmed cases have an obvious continuous distribution trend,mainly concentrated near Wuhan and economically developed cities(e.g.,Shanghai and Beijing),while the areas with fewer cases have relatively stable coverage(e.g.,Qinghai,Tibet,and Xinjiang),indicating that regional differences are increasing.(2)From the global Moran’s I index,from January 31 to September 1,the accumulative confirmed cases at the prefecture-level showed significant global spatial autocorrelation(P <0.0001,Z> 9.58),indicating that the accumulative confirmed cases at the prefecture-level showed extremely significant spatial dependence.In addition,the results of local spatial autocorrelation analysis show that Hubei province and its neighboring cities belong to the "high-high" cluster region;The "low-low" clustering regions showed a continuous layout trend,and these regions were mainly located in Inner Mongolia,Gansu,Ningxia,Qinghai,Tibet,and Xinjiang,while the rest of the administrative regions were not statistically significant.(3)Hotspot analysis(Getis-Ord G *)shows that 90% of hotspots spread from 32prefecture-level cities in January to 67 prefecture-level cities in March.There were 26 hotspots of great importance(99% Confidence)and 41 hotspots of great importance(95% Confidence),which gradually stabilized.(4)The results of Spatio-temporal scanning showed that a total of 6 clusters were detected,including 17 provinces.The higher the Risk Ratio(RR),the higher the risk of disease in this region.P<0.05 rejected the null hypothesis and was statistically significant.(5)Social factors(population migration scale index,government response severity index)and natural factors(air pollutants and meteorological factors),and daily confirmed cases of COVID-19 were analyzed.The results show that factors such as average daily temperature,wind speed,precipitation,air pollution concentration,migration scale index,and government response severity index are closely related to the incidence of COVID-19.(6)The ARIMA and LSTM time series models incorporating risk factors were used to model and evaluate the incidence of COVID-19 in China.Through comparison,it is found that LSTM risk-fitting assessment model has a lower effect than the ARIMA risk-fitting assessment model,that is,the LSTM model is better than the ARIMA model.Finally,models with relatively high accuracy were used to predict the spread of COVID-19 in China.The results show that the LSTM risk propagation prediction model achieves better modeling and evaluation effect.In this study,the number of confirmed COVID-19 cases in China was taken as the measurement index,and a variety of research methods were adopted to comprehensively explore the Spatio-temporal characteristics,distribution rules,and transmission risks of COVID-19 transmission in China from multiple perspectives,providing an information-based decision-making tool for the management of COVID-19 transmission by public health departments.