A novel hybrid fuzzy time series model for prediction of COVID-19 infected cases and deaths in India. (May 2022)

• Record Type: Journal Article
• Authors: Kumar, Niteesh; Kumar, Harendra
• Is Part Of: ISA transactions. Volume 124(2022)
• Page Start: 69
• Page End: 81
• Publication Date: 2022-05
• Subjects: COVID-19 -- Fuzzy C-means -- Fuzzy time series -- Clustering -- Pandemic; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2021.07.003
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: World is facing stress due to unpredicted pandemic of novel COVID-19. Daily growing magnitude of confirmed cases of COVID-19 put the whole world humanity at high risk and it has made a pressure on health professionals to get rid of it as soon as possible. So, it becomes necessary to predict the number of upcoming cases in future for the preparation of future plan-of-action and medical set-ups. The present manuscript proposed a hybrid fuzzy time series model for the prediction of upcoming COVID-19 infected cases and deaths in India by using modified fuzzy C-means clustering technique. Proposed model has two phases. In phase-I, modified fuzzy C-means clustering technique is used to form basic intervals with the help of clusters centroid while in phase-II, these intervals are upgraded to form sub-intervals. The proposed model is tested against available COVID-19 data for the measurement of its performance based on mean square error, root mean square error and average forecasting error rate. The novelty of the proposed model lies in the prediction of COVID-19 infected cases and deaths for next coming 31 days. Beside of this, estimation for the approximate number of isolation beds and ICU required has been carried out. The projection of the present model is to provide a base for the decision makers for making protection plan during COVID-19 pandemic. Highlights: A hybrid fuzzy time series based model is proposed. FCM clustering is modified by using an exponential function to tolerate noisy data. Present model is able to predict approximate COVID-19 infected cases and deaths. The model presents prediction of COVID-19 infected cases and deaths for next 31 days.