Robust geographically weighted regression: Modeling sugarcane yield in East Java

Geographically weighted regression (GWR) is a popular method for examining spatial heterogeneity in a regression model. However, this method is inherently unreliable for outliers, which can lead to a biased estimate of the underlying regression model. Robust Geographically Weighted Regression (RGWR) is a method that extends the GWR model to handle outliers in the dataset. RGWR is a robust regression method that uses a weight function to reduce the effect of outliers in the dataset RGWR handles outliers in the data by down-weighting their influence using a weighting function such as tukey bisquare. In this study, RGWR will be applied to sugarcane yield data in East Java. Sugarcane is major crop in this region, and modeling its yield. It can help farmers predict how much they will be able to harvest each season. This information can then be used to make better decision about planting, fertilizing, and harvesting. Additionally, accurate yiels models can help ensuring food security in the region. In short, modeling sugarcane yield is a crucial step towards building a more sustainable for east java and can reduce sugar import. Based on the AIC, MSE and R2, it is known that in data containing outliers, the performance of RGWR is better than GWR.