A Research On Prediction And Risk Assessment Of COVID-19 Development Trend

Since the outbreak of the novel coronavirus 2019(COVID-19),the virus has ravaged most countries and regions around the world,with459,345,074 confirmed cases and 6,045,258 deaths as of March 15,2022.The virus has had a significant impact on health care systems,economies and societies in various countries and regions.With the outbreak of the virus to the continuous rapid spread,how to scientifically analyze and predict the development trend of the epidemic is an important research problem.In this thesis,we first propose a time-window-based SIR prediction model(TW-SIR)for early prediction of infectious disease outbreaks based on traditional infectious disease models,which dynamically assesses the basic number of infections during the spread of an outbreak and the growth rate of the outbreak index based on a time window,and solves and predicts the model using the fourth-order Runge-Kutta method.Then as the infectious disease continues to develop and the amount of available data increases,we propose a SIRVD-DL model based on time-varying SIRVD and deep learning for short-term and medium-term prediction,which combines a SIRVD model with vaccination status and a deep learning approach to make the prediction more accurate,as well as more interpretable and robust.Finally,we propose a COVID-19 risk assessment framework that integrates hazard,exposure,vulnerability,and resilience of a country and region to calculate its risk index,and achieves long-term prediction of transmission risk and mortality risk based on TFT.The experimental results on public datasets in this thesis show that the proposed time-window-based SIR prediction model can effectively assess and measure the real-time changes of epidemic parameters during the transmission of infectious diseases,including the basic transmission number and the index growth rate,and is adapted to the early prediction of different infectious diseases;the proposed time-varying SIRVD-based deep learning prediction model can assess the metric transmission,cure and mortality rates in the context of vaccination The proposed COVID-19 risk assessment framework can effectively assess the risk factors in each country and region,and the TFT-based long-term risk prediction model can effectively predict the cumulative confirmed diagnosis and cumulative death.