GSTAR (1;1) Transfer Function Model for Forecasting Chili Prices with Rainfall Effect

Yundari, Yundari (Department of Mathematics, Universitas Tanjungpura, Pontianak, Indonesia)
Rahmawati, Asri (Department of Mathematics, Universitas Tanjungpura, Pontianak, Indonesia)
Pratiwi, Yuyun Eka (Department of Mathematics, Universitas Tanjungpura, Pontianak, Indonesia)

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

Chili price fluctuations are strongly influenced by climate variability, distribution inefficiencies, and spatial. The ARIMA–TF and GSTARX models can address these issues. However, GSTARX is generally limited to linear responses of exogenous variables, while ARIMA–TF does not capture spatial heterogeneity. This study proposes an integration of the GSTAR and Transfer Function approaches. This model combines spatial-temporal dependencies with the external factors such as rainfall. The modeling process begins with GSTAR modeling, followed by parameter estimation of the transfer function using the Marquardt algorithm. The data used are rainfall and chili prices in three locations in West Kalimantan. The model data estimation results show very good accuracy, with MAPE values of 15%, 8%, and 14%. These results confirm that rainfall is a crucial factor influencing chili price variations and demonstrate the reliability of GSTAR–TF in forecasting agricultural commodities in the face of climate variability.