Mixed Geographically Weighted Regression Modeling on Gross Regional Domestic Product of Districts/Cities in Central Java

Mixed Geographically Weighted Regression (MGWR) is a combined model between multiple linear regression models and GWR models. So that the Mixed GWR model will produce global parameter estimates and other parameters are local according to the location of observation. The purpose of this study is to form a Mixed Geographically Weighted Regression (MGWR) model with the best weights formed in modeling district / city GRDP data in Central Java and identify variables that significantly affect district / city GRDP in Central Java. Data on HDI, Percentage of Poor Population, TPT, TPAK, and UMR are used as predictor variables to explain district/city GRDP in Central Java obtained through the website of the Central Java BPS publication. Modeling uses the best weight obtained through the minimum Cross Validation (CV) value, namely the fixed tricube kernel function weight. The results of the model formed using the Mixed Geographically Weighted Regression (MGWR) method with the selected optimum weighting function fixed tricube are different for each district / city in Central Java and produce variables of the percentage of poor people, the open unemployment rate, the labor force participation rate and the regional minimum wage have a localized nature of a region that is significant to the model, then there are no global variables that are significant to the model.