Abstract

In this paper we present a report on the performance of two black-box mathematical models of COVID-19 prediction. The models are the growth equation models: modified Verhulst Logistic Model and the Morgan-Mercer-Flodin (MMF) Model. Both models predict the cumulative numbers of infected people using the COVID-19 Data of Indonesia from begining day of the infection to end of September 2020. The COVID-19 pandemic data in Indonesia from the Worldometer website (Worldometer 2020) will be used as the underlying data for the curve fitting to those two models. We use the time series data of the total confirmed cases and the daily new cases to fit to the model models. We used the data starting on 2 March, the official first day of the reported pandemic cases in Indonesia, up to 30 November 2020. We used the models to fit the time series of the total confirmed cases by minimizing the minimum root of the mean square error (RMSE) to obtained the parameterized models. After a year has been lapped, we compare the prediction from those parameterized models to the actual number of infected people for the same country, from the beginning of epidemic to 30 August 2021. The results show that for certain condition (different assumptions of the two models) the models could predict the number in a relatively good accuracy, at least qualitatively could portray the shape of the cumulative number of infected people curve.