Application Research Of Intelligent Prediction Of Novel Coronavirus Pneumonia Based On Machine Learnin

Since December 2019,the COVID-19 has had a serious impact on every country in the world.Due to the relatively poor medical conditions,insufficient medical equipment,and overloaded hospital capacity in some countries and regions,medical staff cannot quickly diagnose and screen COVID-19,the research and development of drugs and vaccines lags behind,and the development trend of the epidemic is not clear,resulting in most patients being unable to receive immediate treatment,further aggravating the COVID-19 pandemic.As a powerful tool,Artificial Intelligence technology represented by Machine Learning can help humans solve the above-mentioned complex problems.Therefore,this paper focuses on the application of Machine Learning in the intelligent prediction of COVID-19.The main work content and innovations are as follows:(1)In response to the problem that medical staff could not quickly determine the severity of admitted patients during the pandemic,this study developed an algorithm that can predict the severity of patients with high accuracy using only lightweight Artificial Neural Network(ANN),Recurrent Neural Network(RNN)and ten binary features.In this study,public data from clinics and hospitals in the public health system of Sonora and Tlaxcala in Mexico were used as datasets,13 commonly used machine learning algorithms were compared,and it was found that ANN and RNN performed best,and Shapley additive explanation(SHAP)was used for interpretability analysis.Finally,the knowledge distillation model was used to lightweight the model and features,and an algorithm with an area under the receiver operating characteristic curve(AUC)of 0.87 was trained.The algorithm can quickly and accurately predict the severity of patients,solve the problem of large model size and large input data,facilitate deployment on smart devices,and reduce the pressure of medical staff.(2)In this study,the COVID-19 pandemic trend prediction model combining a mathematical model of infectious diseases and a deep learning model was developed in view of the problem that the COVID-19 epidemic was difficult to control and the current model could not make scientific predictions of the future development trend of the epidemic.In this study,the epidemic data set released by Johns Hopkins University in the United States was used as the research sample,and the mathematical model of infectious diseases was used to replace the random noise in the Generative Adversarial Network,and the data amplification was carried out,which solved the problem that the training dataset was too small.A comparative study found that the use of SEIRD(Susceptible,Exposed,Infected,Recovered,Dead)infectious disease model instead of random noise was better than other mathematical models of infectious diseases.In this study,also compared the effects of deep learning models,time series prediction models,and mathematical models of infectious diseases,and found that the Transformer model trained by SEIRD model replaces the random noise of the generative adversarial network for data amplification has the best effect,with the Mean Squared Error(MSE)of 0.0008.The model combines the transmission characteristics of infectious diseases and the advantages of deep learning models to scientifically predict the development trend of the epidemic.