COVID-19 Case Growth Prediction Using a Hybrid Fuzzy Time Series Forecasting Model and a Machine Learning Approach

Abstract

The COVID-19 pandemic has evolved into a global health crisis, with Indonesia particularly affected due to its high death rates compared to the rest of Asia. A significant number of unacknowledged, undocumented, or unaddressed cases further exacerbate the situation in Indonesia. Challenges arise from the growing patient population and a scarcity of resources, medical experts, and facilities. This study analyzes the daily development of COVID-19 cases in Indonesia, aiming to estimate the number of confirmed cases, recoveries, and fatalities. Introducing a novel hybrid forecasting model, we utilize the Holt-Winter triple exponential smoothing statistical method and the fuzzy time series rate of change algorithm. We apply the Triple Exponential Smoothing Holt Winter statistical model to predict future periods to the fuzzy time series. Based on the testing results, our proposed hybrid forecasting model demonstrates a very high level of predictive capacity. The acquired data are highly accurate, with a 0.15 percent confirmation rate, 0.15 percent recovery rate, and a 0.20 percent mortality rate, along with an average absolute error of less than 10% for each COVID-19 case. The findings indicate that early awareness by the COVID-19 Task Force of the status of cases is highly advantageous. This awareness can aid in formulating appropriate policies for future planning, organization, and accelerated treatment of COVID-19 in Indonesia. Consequently, successful efforts can be made to slow the emergence and spread of COVID-19 in the country.