Fig. 4.

Forecast of COVID-19 epidemics in Heilongjiang province (data for first 18 days of 38 in total are used for modeling fitting (training set), while the rest 20 data points are used for validation (test set)), Tianjin (17/39), Hunan province (15/39), Guangdong province (15/40), Chongqing (15/39) and Xiaogan city (18/37) in Hubei province (central, city with the second largest reported infected populations). The upper panel (A) shows the results of five empirical functions, the middle one (B) for four different statistical inference methods combined with the Logistic function (the exponent $\gamma$ derived from $R_0$), while the lower panel (C) gives the results of five ODE models. The one with the smallest RMSE to the training data is drawn.