Analysis And Prediction Of New Coronary Pneumonia Infection Cases

At the end of 2019,the novel coronavirus attacked the human race,it was highly infectious and could cause serious symptoms or even death of the patients,which poses great challenges to its prevention and control.According to real-time data of COVID-19,we analyzed the trend,spatial distribution and propagation characteristics of the epidemic,and the results can help the government adjust the relevant prevention and control policies in time to prevent and control the epidemic.The spatial distribution of confirmed cases of COVID-19 in 164 countries was explored through spatial clustering analysis and Kriging interpolation.It was observed that except for the spatial clustering of the outbreak in the early April 2020,the outbreak was random.Over the past year,the number of confirmed cases of COVID-19 has been increasing and expanding.Based on Person correlation analysis and Lasso regression model,the relationship between economic development level,medical level,prevention and control efforts and infection rate of COVID-19 was analyzed.The results showed that there was a positive correlation between economic development level and infection rate,and a negative correlation between prevention and control efforts and infection rate.The estimation of effective reproduction number is helpful to determine the transmission capacity of the virus after different intensity of prevention and control,and then to study the impact of the implementation of epidemic prevention and control policies on the control of the epidemic.In this dissertation,a mathematical model of effective reproduction number is constructed and the effective reproduction number is estimated based on the maximum likelihood method to solve this problem,and then the changes of effective reproduction number are calculated in different periods of time in164 countries,the prevention and control effect of epidemic situation are analyzed in different countries under different intensity of prevention and control policies.Finally,XGBoost model,LSTM model and Prophet model were constructed respectively to model and predict the daily newly confirmed cases in major epidemic countries,and the prediction effect of each model was compared by error.In order to make each model give full play to its advantages,XGBoost model,LSTM model and Prophet model are weighted and optimized by the method of reciprocal variance,and then the daily number of newly diagnosed cases was predicted by using the combined model.The case analysis shows that XGBoost-LSTM-Prophet combination model has more accurate prediction results.