Computational framework for smart tourism management: hybrid time series decomposition and predictive modeling

Abstract

Smart tourism management in rural multi-destination settings requires forecasting methods that are accurate enough to support visitor allocation, infrastructure readiness, and ecological protection. This study presents a decomposition-based forecasting framework for Sidowayah Village, Central Java, Indonesia, which integrates three attractions with different demand profiles: Umbul Manten, Siblarak, and Kampung Dolanan. Using monthly visitation data from May 2023 to April 2024, the study compares additive and multiplicative decomposition models within a common workflow of data collection, preprocessing, trend-seasonal decomposition, model evaluation, and sustainability-oriented interpretation. The contribution of the study lies in clarifying destination-specific criteria for selecting additive versus multiplicative models, improving methodological transparency in preprocessing and temporal validation, and translating forecast outputs into practical smart tourism actions aligned with sustainable development goals (SDGs) 11 and 12. The results show that the multiplicative-average all model yields the lowest mean absolute percentage error (MAPE) for Umbul Manten (14.1%) and Siblarak (56.8%), while the additive-centered moving average model is more suitable for Kampung Dolanan based on mean absolute deviation (MAD) (162.6). Although the 12-month dataset limits long-term generalization, the framework provides a reproducible basis for data-informed tourism management in rural destinations.