A mathematical model and forecast for the coronavirus disease COVID-19 in South Korea
(A.B. Alyokhin, B.V. Burkynskyi, A.N. Grabovoi, V.A, Dilenko, N.I. Khumarova)
14.08.2020. No 141.
On June 15, 2020, we published the latest predictive monitor (see https://www.facebook.com/ab.alyokhin/posts/152284093081161) for the COVID-19 coronavirus outbreak in South Korea. This forecast based on statistical data from the start of observations inclusive June 14, 2020.
We have regularly monitored the accuracy of this forecast:
in addition, every time he pleased us with his accuracy.
However, the time has come to say goodbye to him. Moreover, we will do this after 60 days from the moment of its compilation with a short final analysis of its accuracy, the quantitative indicators of which are giving in table. 1–2 fig. 1–2.
Table 1 (Fig. 1) shows the calculated and actual values of the main indicators of the epidemic in South Korea for the entire forecast period, the absolute and relative errors of daily forecasts, as well as the average absolute error MAE and the average absolute error in percentage MAPE of the entire forecast as a whole.
Table 2 shows the same data on case fatality rates I (TC), I (CC) and the IP progress indicator.
In both tables, taking into account their size, there are hidden lines that were included in the materials of the latest monitors devoted to the analysis of the accuracy of this forecast. The days that have passed since the last such publication are highlighting in purple.
The data in these tables, including the error estimates, indicate that the 60-day forecast is still highly accurate.
The mean absolute percentage errors (MAPE * estimates) of the forecast for the main indicators of the epidemic were as follows:
- total number of infected people — 3.76%;
- total number of deaths — 1.92%;
- total number of recovered — 1.97%.
The errors in forecasting the most important derived indicators of the coronavirus epidemic in Ukraine (MAPE estimates) are as follows (Table 2, Fig. 2):
- mortality rate I (TC) — 4.33%;
- mortality rate I (CC) — 3.50%;
- IP progress indicator — 4.40%,
Recall that the accuracy (according to the MAPE accuracy criterion) is usually consider high with an error of up to 10%, good — from 10% to 20% and satisfactory — from 20% to 50%.
Diagrams with the actual and projected trajectories of the main and most important derived indicators of the COVID-19 coronavirus epidemic in South Korea for a forecast period of 60 days differ little from those that given in the previous analytical review (see https://www.facebook.com /ab.alyokhin/posts/170624467913790), containing estimates of the accuracy of this forecast over a forecast period of 55 days. Therefore, such diagrams are not showing in this monitor.
However, according to the table. 1–2 it is already clearly seen (see the level of daily relative errors of MAPE)) that this forecast has approached the exhaustion of its forecasting capabilities and needs to be update. In addition, it would undoubtedly be updating in a timely manner if there were a social order for it.
Since, when predicting the development of the coronavirus epidemic in different countries, we are guiding, first, by scientific interests in the field of improving mathematical models and forecasting methods, we allow ourselves to test various strength predictions, including the use of extreme lead periods.
The preservation of the high accuracy forecast for such a long time indicates the quality of the model and the forecast itself, the stability of the trends in the development of the coronavirus epidemic in South Korea and very low, in contrast to many European countries and other regions, the second wave associated with the weakening of quarantine. This explains the slow, almost linear increase in the error of this forecast.
However, South Korea still failed to avoid losses associated with the weakening of quarantine. One of the most innocuous victims of this was our forecast, which 2 months ago could not have assumed that South Korea would be so effective, for example, to the whole world, taking control of the epidemic in its acute phase, it will relax somewhat near the finish line.
Sources of statistical data:
Our materials also:
Accuracy of our forecasts:
Publications on case fatality rates and progress indicator