Time series prediction of COVID-19 transmission in America using LSTM and XGBoost algorithms

Luo, Junling; Zhang, Zhongliang; Fu, Yao; Rao, Feng

In this paper, we establish daily confirmed infected cases prediction models for the time series data of America by applying both the long short-term memory (LSTM) and extreme gradient boosting (XGBoost) algorithms, and employ four performance parameters as MAE, MSE, RMSE, and MAPE to evaluate the effect of model fitting. LSTM is applied to reliably estimate accuracy due to the long-term attribute and diversity of COVID-19 epidemic data. Using XGBoost model, we conduct a sensitivity analysis to determine the robustness of predictive model to parameter features. Our results reveal that achieving a reduction in the contact rate between susceptible and infected individuals by isolated the uninfected individuals, can effectively reduce the number of daily confirmed cases. By combining the restrictive social distancing and contact tracing, the elimination of ongoing COVID-19 pandemic is possible. Our predictions are based on real time series data with reasonable assumptions, whereas the accurate course of epidemic heavily depends on how and when quarantine, isolation and precautionary measures are enforced.