A Best Exponential Smoothing Method With Hyperparameter Tuning to Predict the Number of Pandemic Cases

The SARSCov-2 virus or Corona Virus is the cause of the emergence of Covid-19 Pandemic (Corona Virus Disease 2019). Since the entry of Covid-19 in Indonesia, prevention efforts have been carried out by the Government through PSBB rules to multilevel PPKM, efforts to predict are also carried out by researchers to determine the number of additional cases in the future. This study aims to determine the best model in predicting the number of COVID-19 cases in Indonesia. The source of this research data is a time series dataset of COVID-19 cases in Indonesia from January 2020 to August 2022. Researchers used the Exponential Smoothing Method with hyperparameter tuning to perform prediction calculations. The results showed that the Single Exponential Smoothing method had a MAPE value of 5.52%, the Double Exponential Smoothing method had a MAPE value of 0.74%, while the Triple Exponential Smoothing method had a MAPE value of 8.60%, so it can be concluded that Double Exponential Smoothing has good performance, while Triple Exponential Smoothing has poor performance in making predictions.