Prediction of spread trend of epidemic based on spatial-temporal sequence

Inhalt:
Effective prevention against epidemic is a common challenge facing all mankind. Spatial-Temporal Attention Graph Convolutional Networks (STAGCN) based on spatial-temporal sequence is proposed to predict the spread trend of epidemic. Firstly, considering the spatial-temporal correlation of epidemic data, the model replaces the graph convolutional network with the graph attention network, meanwhile, utilizing the attention mechanism to adaptively allocate the weights of the epidemic data at different time steps. It increases the weights of important feature, greatly retains the key information of spatial-temporal sequence. Therefore, the ability to capture the spatial-temporal relationship of epidemic data is improved. Secondly, weights of edges are introduced to represent the correlations between target nodes’ characteristics and adjacent nodes’ characteristics, thus, the ability to capture the spatial information of epidemic is improved. Eventually, the experiment result shows that STAGCN model can accurately predicts the spread trend of epidemic in the future, which gives the predicted values of new infections in different regions, providing a reference for the decision- making of epidemic prevention.