ABSTRACT
This paper presents a deep learning approach to predict new COVID-19 infected cases in a specific country with insufficient data at the onset of the outbreak. We collected data on daily new confirmed cases in several countries of the region where COVID-19 occurred earlier and caused more severe effects than in Vietnam. Then we computed some deep machine learning models to adapt the spreading speed of Delta strain in each nation to generate various scenarios for the epidemic situation in Vietnam. We used models based on recurrent neural networks (RNN) architectures such as long-short term memory (LSTM), gated recurrent unit (GRU), and several hybrid structures between LSTM and GRU. Learning from the experiments in this research, we built a set of circumstances for COVID-19 in Vietnam. We also found that GRU always gives the best performance in terms of MSE, while LSTM is the worst. © 2021, Springer Nature Singapore Pte Ltd.