Forecasting the Spread of COVID-19 Based on Differential Transfer Graph Neural Network

The pandemic outbreak of the 2019 Coronavirus disease (COVID-19) has a significant effect on the health and lives of people around the world. Accurately predicting the spread of the disease is vital for fighting the disease during a pandemic. This paper presents a prediction model named Differential Transfer Graph Neural Network (DTGNN). The proposed model takes the differential information of time series into account, which includes the important trend of disease evolution. Since the limited training data may cause the phenomenon of underfitting, we employ transfer learning method to train the model. We first learn the model for countries with disease widespread, and then transfer it to the target country where disease is in the early stages. Compared with the baseline models, the proposed model performs better in mid- and long-term predictions.