Forecasting Development of COVID-19 Epidemic in European Union Using Entropy-Randomized Approach

The paper is devoted to the forecasting of COVID-19 epidemic by novel method randomized machine learning. This based on idea estimation probability distributions model parameters and noises real data. Entropy-optimal correspond state maximum uncertainty which allows resulting forecasts be used as most "negative" scenario process under study. estimates noises, are distributions, must generated, thus obtaining an ensemble trajectories that considered analyzed statistical methods. In this work, for purposes such analysis, mean median over calculated, well trajectory corresponding distribution values parameters. proposed approach predict total number infected people using a three-parameter logistic growth model. conducted experiment data in several countries European Union. main goal demonstrate entropy-randomized predicting near peak. significant contained available modeled additive noise within 30%, both at training stages. To tune hyperparameters model, scheme configure them according testing dataset with subsequent retraining It shown same datasets, makes it possible development more efficiently comparison standard least-squares method.