Implementation of Geographically and Temporally Weighted Regression with Cross Validation and Generalized Cross Validation Methods for Deforestation Modeling in Kalimantan

J Statistika : Jurnal Ilmiah Teori Dan Aplikasi Statistika
Vol 18 No 1 (2025): Jurnal Ilmiah Teori dan Aplikasi Statistika

Deforestation in Indonesia has received national and international attention for its ecological, economic and social impacts. Kalimantan is an area with high deforestation rates triggered by various factors that vary between locations and time. This study aims to model the deforestation rate in Kalimantan during the period 2014 to 2022 using the Geographically and Temporally Weighted Regression (GTWR) method. The model was tested with a combination of Fixed and Adaptive Gaussian Kernel weighting functions and Cross Validation (CV) and Generalized Cross Validation (GCV) bandwidth determination methods. The results show that the best model is GTWR with Fixed Gaussian Kernel and GCV based on R2 and AIC values. Spatial-temporal analysis shows that neither variable is significant in 2014, forest fires are significant in 2019, and in other years both variables are broadly significant. The findings provide insights into the spatial and temporal dynamics of deforestation factors to support area-based policies.