Prediksi Produktivitas Padi Menggunakan Geographically Weighted Regression (GWR) di Kabupaten Jember

Abstract

Rice productivity plays an important role in supporting food security and sustainable agricultural development, in line with Sustainable Development Goals 2 (SDGs 2), particularly the goal of ending hunger, improving food security, and promoting sustainable agriculture. Rice productivity in Jember Regency shows variations between sub-districts due to differences in regional characteristics. This study aims to analyze the spatial distribution of rice productivity in 2025 and compare the capabilities of Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR) models in explaining the relationship between harvested area, rainfall, slope, and temperature on rice productivity. The study used secondary data from 31 sub-districts in Jember Regency. The analysis was conducted through spatial mapping, descriptive statistics, OLS modeling, and GWR modeling using an Adaptive Bisquare Kernel with an optimum bandwidth of 28 neighbors. The results showed that rice productivity ranged from 57.00 to 64.50 quintals per hectare. The OLS model produced an R² value of 0.3655, an AICc of 135.8336, an RMSE of 1.69, an MAE of 1.31, and a MAPE of 2.15%. Meanwhile, the GWR model produced an R² value of 0.5096, an AICc of 96.5934, an RMSE of 1.48, an MAE of 1.12, and a MAPE of 1.83%. These results indicate that GWR provides a better model fit and lower estimation errors than OLS because it is able to describe spatial variations in the relationships between variables across sub districts. The findings are expected to provide spatial information that can support location-specific agricultural planning and efforts to improve rice productivity in accordance with the objectives of SDGs 2.