PEMODELAN DEFORESTASI HUTAN LINDUNG DI INDONESIA MENGGUNAKAN MODEL GEOGRAPHICALLY AND TEMPORALLY WEIGHTED REGRESSION (GTWR)

*Thea Zulfa Adiningrumh  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Alan Prahutama  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Rukun Santoso  -  Departemen Statistika, Fakultas Sains dan Matematika, Universitas Diponegoro, Indonesia
Received: 13 Feb 2020; Published: 13 Feb 2020.
View
Open Access Copyright 2020 Jurnal Gaussian
License URL: http://creativecommons.org/licenses/by-nc-sa/4.0

Citation Format:
Abstract

Regression analysis is a statistical analysis method that is used to modeling the relationship between dependent variables and independent variables. In the linear regression model only produced parameter estimators are globally, so it’s often called global regression. While to analyze spatial data can be used Geographically Weighted Regression (GWR) method. Geographically and Temporally Weighted Regression (GTWR) is the development of  GWR model to handle the instability of a data both from the spatial and temporal sides simultaneously. In this GWR modeling the weight function used is a Gaussian  Kernel, which requires the bandwidth value as a distance parameter. Optimum bandwidth can be obtained by minimizing the CV (cross validation) coefficient value. By comparing the R-square, Mean Square Error (MSE) and Akaike Information Criterion (AIC) values in both methods, it is known that modeling the level of deforestation in protected forest areas in Indonesia in 2013 through 2016 uses the GTWR method better than global regression. With the R-square value the GTWR model is 25.1%, the MSE value is 0.7833 and AIC value is 349,6917. While the global regression model has R-square value of 15.8%, MSE value of 0.861 and AIC value of 361,3328.

 

Keywords : GWR, GTWR, Bandwidth, Kernel Gaussian

Keywords: GWR, GTWR, Bandwidth, Kernel Gaussian

Article Metrics: