Machine learning models for predicting the number of COVID-19 patients in Ukraine and India
Main Article Content
Models for predicting the number of patients with COVID-19 using machine learning methods have been built. The data for models are obtained from various official sources, including the World Health Organization, from the beginning of the epidemic to the present time. The data in Ukraine and India were selected to teach models for predicting the number of patients with COVID-19. Algorithms of linear regression for Ukraine and gradient boosting for India proved to be the methods that provided high accuracy of the forecast for the existing data. Data analysis was performed using the Python programming language with Sklearn library which is based on SciPy (Scientific Python). In addition, the XGboost gradient boost algorithm library was used. To develop the model, multifactor prediction of time series with the delays as predictors was chosen. It is established that the such characteristics as the date of the event, day of the week, week number, month affect to the model. Model errors are smallest and forecast accuracy were estimated with the best values of 0.83 for Ukraine and 0.75 for India. The built models allow to predict the epidemiological situation in the future, to coordinate actions in different areas of health care and to carry out reasonable preventive measures at the state level.
(2019), Coronavirus disease 2019 COVID-19 (wikipedia.org), available at: https://en.wikipedia.org/wiki/COVID-19.
(2020), Artificial Intelligence and Machine Learning Trends in 2020, available at: https://www.dataversity.net/artificial-intelligence-and-machine-learning-trends-in-2020/Artificial Intelligence and Machine Learning Trends in 2020 - DATAVERSITY
Kononova, K.Yu. (2020), Machine learning: methods and models, V.N. Karazin KhNU Kharkiv, 301 p.
Goodfellow, J., Curville, A. and Bengio I. (2018), Deep Learning, 654 p.
(2021), CSSEGIS and Data/COVID-19, available at: https://raw.githubusercontent.com/CSSEGISandData/COVID-19/master/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv
(2021), GeneXproTools 5.0 – Relative Absolute Error, available at: https://www.gepsoft.com/GeneXproTools/AnalysesAndComputations/MeasuresOfFit/RelativeAbsoluteError.htm
(2021), GeneXproTools 5.0 – Relative Squared Error, available at: https://www.gepsoft.com/GeneXproTools/AnalysesAndComputations/MeasuresOfFit/RelativeSquaredError.htm