<![CDATA[Rainfall modeling often involves complex spatial patterns that vary across locations. Traditional spatial models such as Geographically Weighted Regression (GWR) assume linear relationships and may fall short in capturing nonlinear interactions among predictors and the small sample size is more challenging to fix the assumptions. To address this limitation, this study applies the Geographically Weighted Random Forest (GWRF) method is a hybrid approach that integrates Random Forest (RF), a non-parametric machine learning algorithm with geographically weighted modeling. GWRF is advantageous as it accommodates both spatial heterogeneity and nonlinear relationships, making it suitable for modeling monthly rainfall, which is inherently spatially varied and influenced by complex factors. This study aims to implement and evaluate the performance of the GWRF model in monthly rainfall prediction across East Java. The model is tested using various numbers of trees to determine the optimal structure, and its performance is assessed using Root Mean Square Error (RMSE), Akaike Information Criterion (AIC), and corrected AIC (AICc). Results indicate that the model tends to overestimate the Out-of-Bag (OOB) Error at all tree variations, with the smallest RMSE (85.68) achieved at 750 trees. Humidity emerges as the most influential variable in predicting monthly rainfall in the region, based on variable importance analysis]]>
Copyrights © 2025
