Abstract:
The aim of this study is to contribute to the literature by estimating the 5-weeks number ofcases/deaths for each continent by using statistical-based prediction models, which are quiteeffective on simple but small-scale datasets. While Auto.arima, Tbats, Naive, Holt, Thetaf and, Driftmodels were used for prediction processes root mean square error (RMSE), mean absolute error(MAE), and mean absolute percent error (MAPE) metrics were used for evaluating estimates. According to the confirmed cases MAPE metric values of the 5 continents analyzed, the bestpredictions for Asia, Africa, Europe, America, and Oceania were done by Thetaf, Naive, Thetaf,Auto.arima, and Auto.arima models, respectively. The use of very limited data for time seriesestimates such as 57-weeks in the estimation process was a disadvantage. Most models require atleast two cycles, 104 weeks of data, to run. Therefore, we could not use models such as neuralnetwork autoregressive, multilayer perceptrons, extreme learning machines.The results obtained with the prediction models used in this study aim to make more accuratedecisions for the authorized persons dealing with health to be more prepared for future conditionsand health systems.
