The Prediction of the Epidemic Trend of COVID-19 Using Neural Networks

作者：Jing Yang, Zhen Shen, Xisong Dong, Xiuqin Shang, Wei Li, Gang Xiong · 发表于：IFAC-PapersOnLine · 年份：2020 · DOI：10.1016/j.ifacol.2021.04.182 · 被引用次数：8 · 研究领域：COVID-19 epidemiological studies、COVID-19 diagnosis using AI、Anomaly Detection Techniques and Applications

In this paper, a BP neural network and an LSTM network are applied respectively to the prediction of Coronavirus Disease 2019 (COVID-19) in Wuhan, China and South Korea. The methods do not require specific theories of modelling and the predicted values can be obtained as long as the conventional parameters are set. The mean absolute percentage error (MAPE) of all the experiments are below 5% and the values of the determinable coefficient R2 are all larger than 0.9. The experiments show that the models can fit the actual values well and make relatively accurate predictions. As of March 29, 2020, the cumulative number of confirmed cases in Wuhan is expected to reach 50,068 using BP neural networks and 49,972 using LSTM network, respectively. As of April 13, 2020, the cumulative number of confirmed cases in South Korea is expected to reach 8,862 using BP neural networks and 8,716 using LSTM network, respectively. The models of neural networks are effective in predicting the trend of the COVID-19 epidemic, which is meaningful to prevent and control the epidemic.