EVALUATING THE PERFORMANCE OF HOLT-WINTERS AND ARIMA FORECASTING MODELS FOR DOMESTIC TOURIST ARRIVALS IN EAST JAVA

Abstract

Domestic tourist data in East Java constitute an important variable in forecasting, as they represent the pattern of domestic tourism demand over time. In time series analysis, such as the Autoregressive Integrated Moving Average (ARIMA) method or Holt-Winters exponential smoothing, domestic tourist data help capture trend and seasonal patterns occurring in the tourism sector. This study aims to evaluate the performance of the Holt-Winters and ARIMA models using domestic tourist data in East Java for the period 2021 to 2025. Model performance is assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) as goodness of fit criteria. The results indicate that the ARIMA (2,0,2) model produces the lowest MAE and RMSE values compared to the Holt-Winters model, while the Holt-Winters model only yields the lowest MAPE value; therefore, the ARIMA (2,0,2) model is selected as the best model for forecasting the number of domestic tourists in East Java.