Background
The COVID-19 pandemic is one of the most severe global health epidemics in recent decades. Its consequences have affected hundreds of millions of people in countries around the world because of the high contagiousness and mortality rate.
Results
To further improve the prediction accuracy of the long-term spreading trend of COVID-19, this paper proposes a hybrid neural network prediction model based on a bidirectional long short-term memory network (Bi-LSTM) combined with a multi-head self-attention mechanism. To achieve long-term prediction, this model combines multiple linear regression with the improved susceptible-exposed-infected-recovered (SEIR) model. The bidirectional long short-term memory network can mine important features of input data in both forward and backward directions, and the multi-head self-attention mechanism can capture different attention information to improve the expression ability of the model and help improve the prediction performance. The comparative analysis and prediction of multiple models are based on official real data.
Conclusion
The experimental results show that compared with the long short-term memory network (LSTM) and single chamber model, the proposed COVID-19 spreading model can achieve higher prediction accuracy.