Epidemic Case Prediction of COVID-19: Using Regression and Deep based Models

Abstract

COVID-19, as an international concern of public health emergency, carries the property of high death and infection rates. Researchers need to give an accurate prediction of the daily increase in COVID-19. Though the 2002-2003 SARS breakout provides prescient guidance for these issues, there exist two bottlenecks. First, traditional models that are popular during the SARS period are not able to fit the trend of COVID-19 and predict the cases effectively. Second, the worldwide spreading of COVID-19 also causes the traditional model to fail its function. In this study, we apply several regression models and deep based models for prediction of the COVID-19 pandemic. We perform L1-norm to compute feature-selection; besides, we also introduce SIR, SEIR models to improve our model accuracy. Then, we measure the accuracy of models by Mean Squared Error(MSE). This study concludes that the SEIR model is the best model with the highest performance among the tested approaches.