COVID-19 Trend Forecasting by Using Dropout - LSTM Model

To improve the accuracy of COVID-19 trend forecasting, a method of COVID-19 trend forecasting by using dropout and long short-term memory (LSTM) is proposed. The method uses web crawler based on python to obtain complete domestic historical data of COVID-19, which improves the efficiency of data collection and reduces data errors caused by subjective reasons. To avoid adding time features artificially and explore the nonlinear relationship fully between the less data of COVID-19, the proposed model extends the layers of the deep learning network. Then, the dropout technique is applied to the non-circular part of each hidden layer to randomly deactivate neurons, preventing the neural network from overfitting. The experiment demonstrates that the method can predict the number of cumulative confirmed cases, current confirmed cases and recovered cases.