Validation of optimization for COVID-19 case forecasting model with deep learning

Abstract

In this study, COVID-19 cases were forecast. Traditionally, statistical models have been used in this field. Recently, however, machine learning models have been applied. Therefore, it is necessary to compare both. 5 models, ARIMA, RNN, LSTM, Seq2Seq, and Stacked LSTM, were trained and evaluated with two loss functions, AIC and RMSE, and compared. The results showed that RNN’s RMSE was the smallest with -49.5% reduction to ARIMA, and Seq2Seq’s R2 was the highest with 0.92.