Improving Hierarchical Tourism Forecasting through the ARIMA-OC (ARIMA Based on The Optimal Combination) Method

Abstract

The research explores the ARIMA-OC method (ARIMA based on the Optimal Combination approach) for Hierarchical Forecasting. In this approach, the ARIMA model is used to forecast each individual time series, and the Optimal Combination (OC) technique is applied to merge these initial forecasts into an updated set of predictions. The study compares the ARIMA model with the Exponential Tail Smoothing (ETS) model, with both models being integrated using five different strategies: the Bottom-up approach (BU), the Top-down approach using Forecasted Proportion (TDFP), two Top-down approaches based on Historical Proportions (TDHP1 and TDHP2), and the Optimal Combination approach (OC). To assess how ARIMA-OC performs with small samples, a simulation was carried out, revealing that ARIMA-OC surpasses the other methods according to the MASE metric. Furthermore, non-parametric tests like the Friedman test and the Nemenyi post-hoc test were used to validate the effectiveness of Hierarchical Forecasting.