Hybrid GSTAR-Machine Learning Model for Forecasting Tourists Numbers in Yogyakarta

Sohibien, Gama Putra (Politeknik Statistika STIS, Jakarta, 13330, Indonesia)
Azmi, Annisa Nurul (Politeknik Statistika STIS, Jakarta, 13330, Indonesia)
Sofa, Wahyuni Andriana (Politeknik Statistika STIS, Jakarta, 13330, Indonesia)
Sumarni, Cucu (Politeknik Statistika STIS, Jakarta, 13330, Indonesia)
Prasetyo, Rindang Bangun (Politeknik Statistika STIS)
Putri, Christiana Anggraeni (Politeknik Statistika STIS)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 9, No 2 (2025): Zero: Jurnal Sains Matematika dan Terapan

Tourism management in DI Yogyakarta is vital to ensure tourism benefits local communities. A key challenge lies in the uncertainty and spatial interdependence of tourist visits among neighboring regions. While the GSTAR model captures spatial relationships, its accuracy decreases with outliers, non-linearity, and assumption violations. To overcome these issues, this study integrates GSTAR with machine learning. Using 168 observations of tourist visits across DI Yogyakarta’s regencies/cities (January 2010–December 2023), GSTAR-GLS-XGBoost model achieved 22–34% lower RMSE than other models. Tourist numbers fluctuate greatly, with peaks in May, June, July, and December. Practically, these findings can help local governments and stakeholders optimize resource allocation, plan promotions, and prepare facilities during peak seasons for sustainable tourism management in DI Yogyakarta.