Application of Weighted Fuzzy Time Series Model to Forecast Epidemic Injuries

Abstract

Aims: It is important to predict the amount of COVID-19 injuries. Since the first suspected case of novel coronavirus (2019-nCoV) on December 1st, 2019, in Wuhan, Hubei Province, China, a total of 40,235 confirmed cases and 909 deaths have been reported in China up to February 10, 2020, evoking fear locally and internationally. Here, based on the large amounts of daily publicly available epidemiological data and the need to make an accurate prediction of future behavior requires the definition of powerful and effective techniques capable of inferring random dependency between the past and the future from observations. In this paper, we apply a rewarding model to predict injuries in areas where COVID-19 is, especially in the Arab region. This forecast uses epidemic injuries data from   March 2nd, 2020 to   July 20th, 2020 in Saudi Arabia. Methodology: We propose the use of weighted fuzzy time series techniques (WFTS) and weighted non-stationary fuzzy time series techniques (WNSFTS) to be compared with the classical Auto-Regressive Integrated Moving Average (ARIMA) statistical method. The available data is not a stationary and should therefore be converted first to stationary to forecast it with (ARIMA) and (WFTS) techniques. We do a log transform and differencing on our injuries dataset. Results: When we examine the original data by Dickey-Fuller Test (DFT) to get p-value, we find it is equal to 0.646, it is more than 0.05 which implies the non-stationarity. The mean square error (MSE), the root mean square error (RMSE) and normalization root mean square error (NRMSE), are applied to compare the accuracy of the methods. The results show that WFTS methods give good services for predicting epidemic injuries in the territory by COVID-19. Conclusion: The use of Weighted Non Stationary Fuzzy Time Series (WNSFTS) in forecasting epidemic injuries problem can provide significantly better results because it is able to predict the infected cases at the next time and achieve great predictive accuracy.