Pemodelan Geographically Weighted Logistic Regression dengan Fungsi Adaptive Gaussian Kernel Terhadap Kemiskinan di Provinsi NTT

Abstract

The Geographically Weighted Logistic Regression (GWLR) model is a logistic regression model development that is applied to spatial data from non-stationary processes. This model is used to predict a model of the data set that has a binary response variable which takes into account the spatial factor. This study will discuss the use of the GWLR model using the adaptive weighting function of the Gaussian kernel in a poverty case study in East Nusa Tenggara Province in 2019.The parameter estimation of the Maximum Likelihood Estimation (MLE) method by giving different weights for each observation location. The weight used is the adaptive Gaussian kernel with the optimum bandwidth selection using the Cross-Validation (CV). Based on the results of testing the parameters of the GWLR model with a weighted adaptive Gaussian kernel, it can be concluded that the factors that influence poverty are local and vary in the 22 observation locations, including GRDP per capita, acceptance of smart Indonesian programs, and projected population growth rates, with a classification accuracy rate of 81,82%.