Geographically Weighted Regression Modeling with Fixed Gaussian Kernel Weighted on Spatial Data (Case Study of Food Security in Tanah Laut District of South Kalimantan)

Abstract

Geographically Weighted Regression (GWR) is a regression model that takes into account the spatial heterogeneity effect. In regression models, often there is a relationship between two or more predictor variables is called multicollinearity. Geographically Weighted Lasso (GWL) is a method used to overcome spatial and spatial heterogeneity of local multicollinearity. The purpose of this study establishes the model by using the method of GWL in the case of spatial heterogeneity and overcome local multicollinearity on the issue of food insecurity in Tanah Laut district. Generally, food insecurity in Tanah Laut district is affected by the percentage of the population without access to electricity, the average number of store/grocery shop, and percentage of children under five and maternal mortality. GWL models obtained in accordance with the number of observation locations. The results validate the secondary data showed that the model obtained in the study are in accordance with the actual conditions in the field. Models with fixed weighting Gaussian Kernel is able to predict the eight villages with food security conditions are the same as the secondary data.