MULTI-STEP FORECASTING OF COVID-19 CASES IN EUROPEAN COUNTRIES USING TEMPORAL CONVOLUTIONAL NETWORKS

Authors : Osman Tayfun BİŞKİN

Pages : 117-126

Doi:10.22531/muglajsci.875414

View : 18 | Download : 15

Publication Date : 2021-06-29

Article Type : Research Paper

Abstract :The novel Coronavirus insert ignore into journalissuearticles values(COVID-19); has significantly affected millions of people around the world since the first notification until nowadays. The rapid spread of the virus has dramatically increased the workload of healthcare systems in many countries. Therefore, the need for efficient use of the healthcare system leads researchers to forecast the trend of virus spread. For this purpose, Machine Learning insert ignore into journalissuearticles values(ML); and Artificial Intelligence insert ignore into journalissuearticles values(AI); applications have intensively used to struggle against the coronavirus outbreak. In this study, Temporal Convolutional Network insert ignore into journalissuearticles values(TCN); is applied for modeling the cumulative confirmed COVID-19 cases and forecasting the spread of it in various European countries using time series data. It is also presented that numerical examples for comparing performances of TCN against Long-Short Term Memory insert ignore into journalissuearticles values(LSTM); and Gates Recurrent Unitsinsert ignore into journalissuearticles values(GRU); in terms of computation time, root-mean-square error insert ignore into journalissuearticles values(RMSE);, normalized root-mean-square error insert ignore into journalissuearticles values(NRMSE);, root mean squared log error insert ignore into journalissuearticles values(RMSLE);, mean absolute percentage error insert ignore into journalissuearticles values(MAPE);, and symmetric mean absolute percentage error insert ignore into journalissuearticles values(SMAPE);. Simulation results indicate that the Temporal Convolutional Networks used in this manuscript performs better than other models for forecasting the cumulative confirmed COVID-19 cases.
Keywords : COVID 19, Modeling, Forecasting, Machine Learning, Deep Learning, GRU, LSTM, TCN