Multi-step CNN forecasting for COVID-19 multivariate time-series

The new coronavirus (COVID-19) has spread to over 200 countries, with 36 million confirmed cases as of October 10, 2020. As a result, numerous machine learning models capable forecasting the epidemic worldwide have been produced. This paper reviews and summarizes most relevant for COVID-19. dataset is derived from world health organization (WHO) COVID-19 dashboard, it contains official daily counts cases, fatalities, vaccination use reported by territories, regions. We propose various convolutional neural network (CNN) based such CNN, single exponential smoothing CNN (S-CNN), moving average (MA-CNN), smoothed (SMA-CNN), (MAS-CNN). Here, MAPE MSE are used assess suggested models. frequently compare accuracy across time series different scales. MSE, model must strive total forecast equal entire demand. That is, optimizing seeks create that right on so unbiased. final result shows SMA-CNN outperformed its baselines in both MSE. main contribution this novel approach more accurate base strategy preventing spreads.