China’s novel coronavirus pneumonia team published a research paper on “Journal of Thoracic Disease” published in the Journal of Guangzhou respiratory and Health Research Institute on 2 June, based on SEIR optimization model and AI’s prediction of the outbreak of COVID-19 in China (of) under public health intervention. In this study, the migration data before and after January 23 and the latest covid-19 epidemiological data were integrated into the SEIR model and the epidemic curve was obtained.
The forecast shows that the epidemic in China reached a peak in late February and tended to be flat at the end of April. If the implementation of the control measures is delayed for five days, the scale of the epidemic in the mainland of China is expected to expand to three times; if the control efforts in Wuhan are reduced, Hubei may have a second peak in mid March and continue to the late April. Among them, the optimized model forecast shows that there are 90000 to 120000 cases in the mainland of China by the end of April.
(A) The cumulative number of covid-19 cases predicted by LSTM in China, (b) the number of new covid-19 cases calculated based on the actual data (purple), SEIR model (Orange) and LSTM model (green).
At the same time, the research team also reminded that there is a certain gap between the model prediction and the reality. If the government continues to strictly implement the control policy, improve the diagnosis level and launch the use of drugs, the epidemic scale will be greatly controlled. According to the current situation, Hubei provincial government will continue to maintain strict control, and the possibility of a second peak is relatively small.
The predicted results are based on the control policies before February 9, and the epidemic scale is smaller than that predicted by the lancet and overseas scholars. The data shows that by the end of April, the peak number of confirmed cases (non cumulative confirmed cases) in mainland China is no more than 70000, no more than 52000 in Hubei, no more than 1200 in Guangdong and Zhejiang. At present, the epidemic data are within the prediction range of this paper, which is closer to the real situation.
SEIR optimization model and artificial intelligence model can effectively predict the epidemic trend of new coronavirus, which proves that the public health interventions implemented since January 23 have effectively controlled the development of the epidemic. Strict prevention and control measures such as “early screening” are best implemented by the end of April 2020.