Short-term forecast of COVID-19 spread (16.05.20 - 20.05.20) based on the Back Propagation Neural Network

Date of publication 15.05.2020

The results of the short-term predictive modeling of the number of COVID-19 infection cases in Ukraine and the city of Kyiv for 16.05.20 – 20.05.20 (fig. 1-10; tables 1, 2) were obtained using the Back Propagation Multilayer Neural Network based on the “sliding window” mechanism with 12 neural network training data points.

As we can see in case of Ukraine over the fifteen-day period (06.05.20 - 20.05.20), the nature of the pandemic development becomes clearer and more linear, and reaches a plateau of (400-550) new daily infection cases (fig. 2). According to the predictive modeling carried out by the COVID-19 FORESIGHT project team, such nature of the process development may linger until the last ten-day interval of May 2020, when the pandemic peak is likely to occur. Thereafter the pandemic should start declining (the number of new daily infection cases will be consistently lower than the number of daily recovery cases).

As for the city of Kyiv over the fifteen-day period (06.05.20 - 20.05.20), the pandemic development nature is still more transient. Its volatility will probably decrease, and its character will become more linear in the following week with a plateau of (40-50) new daily infection cases (fig. 5, 6).

The mean absolute percentage error of the above forecast does not exceed MAPE = 1.878 for Ukraine and МАРЕ = 2.3% for the city of Kyiv (fig. 9, 10, tables 3, 4). It should be mentioned that the Back Propagation Neural Network is less sensitive to short-term spikes and surges and allows for a lower forecast error with larger volumes of data (as the overall Ukraine data compared to the city of Kyiv).

 As we can see in fig. 11 and 12 , the number of daily recovery cases in Ukraine approaches the number of daily infection cases (on 14.05.20 this ratio (Δ) was even positive for one day). The project team expects that in the third ten-day interval of May 2020 this ratio Δ may become steadily positive, which will represent the beginning of the pandemic decline.    

Scientific supervisor of the project: Michael Zgurovsky.

Project team: Oleksandr Voytko, Nataliia Gorban, Iryna Dzhygyrey, Bohdan Dudka, Kostiantyn Yefremov, Yuriy Zaychenko, Pavlo Kasyanov, Maria Perestyuk, Іvan Pyshnograiev, Victor Putrenko, Viktor Sineglasov.