ABSTRACT
The COVID-19 pandemic has spread rapidly since 2019. The worldwide uncontrollable outbreak has caused health and economic damage. Multiple deep learning predictable models have been proposed to forecast COVID-19 spread that can help monitor the situation. To improve preciseness of predicted results, we propose multiple time series variables that can be used in LSTM based model to get higher accuracy predicted results for both short and long periods of time. COVID-19 cumulative cases, new cases, 5 days simple moving average of cumulative cases and average of cumulative cases in neighboring countries are added as additional features fed to LSTM model to improve predicted results up to 5% better than the LSTM model without additional features on 7, 14, 21, 28-day prediction. © 2021 ACM.