Information Technology of Time Series Forecasting of New COVID-19 Disease Patients Based on Prophet Model
Keywords
Abstract
The article describes the results of the development of information technology for forecasting the time series of the number of COVID-19 patients. A review of the developments on this subject was carried out and the choice of the Facebook Prophet library as the basis for the developed IT was substantiated. It is proposed to increase the accuracy of the forecasting of the number of new coronavirus patients in the short-term using the selected model.
Mathematical substantiation of the available parallel-serial iterative methods for identifying the parameters of the Facebook Prophet model to take into acco0unt the significant volatility of the number of new patients was proposed. The methods make it possible to determine the hyperparameters of the trend of the series, its seasonal components and anomalies affecting the value of this series. The approach to comparative analysis of the influence of the main incidence trends of regions (or countries) without considering the influence of anomalies and seasonal components has been formalized by creating cartograms that enable to analyze trends in the spread of the disease in a given region.
The architecture of the proposed information technology is developed, and its components are described. A Python software based on the Kaggle platform was created, it implements this technology, the results of its application is compared with the SEIR-U model, developed by the scientists of the National Academy of Sciences of Ukraine, based on the reports of the Working Group on Mathematical Modeling of Problems Related to the SARS-CoV-2 Coronavirus Epidemic, at the National Academy of Sciences of Ukraine according to data, obtained in the period 2020—2022. The comparison, carried out, proved the effectiveness of the proposed information technology.
