PENERAPAN METODE GEOGRAPHICALLY WEIGHTED LASSO PADA KASUS PRODUK DOMESTIK REGIONAL BRUTO JAWA BARAT

• Main Author: Muhammad Luthfi Kasyfurrahman, - (Author)
• Format: Book
• Published: 2020-02-05.
• Subjects: Thesis; NonPeerReviewed
• Online Access: Link Metadata

Summary

Geographically Weighted Lasso (GWL) merupakan gabungan dua metode regresi yaitu Geographically Weighted Regression (GWR) dan Least Absolute Shringkage Selection Operator (LASSO). Kedua metode memiliki kegunaannya masing-masing. GWR merupakan regresi dengan memperhatikan aspek lokasi geografis karena uji heterogenitas spasial tidak terpenuhi. LASSO merupakan sebuah metode regresi untuk mengatasi multikolinearitas pada data. Dari dua masalah secara simultan terdapat dalam satu model regresi maka metode GWL tercipta. Dalam penelitian ini akan menerapkan dan menginterpretasi hasil metode Geographically Weighted Lasso dalam masalah Produk Domestik Regional Bruto di Jawa Barat. Berdasarkan hasil analisis, diperoleh 27 model dari 27 pengamatan dengan nilai R-square metode GWL 0,9472643. Salah satu modelnya yaitu model PDRB untuk kota bandung dengan nilai x_3 sebesar 0,292049, nilai x_8 sebesar 0,640292 dan nilai x_12 sebesar -0,01114. Kata Kunci : Heterogenitas spasial, Multikolinearitas, Geographically Weighted Regression (GWR), Least Absolute Shringkage and Selection Operator (LASSO), Geographically Weighted Lasso (GWL), Produk Domestik Regional Bruto Geographically Weighted Lasso (GWL) is a combination of two regresi method that is Geographically Weighted Regression (GWR) and Least Absolute Shringkage and Selection Operator (LASSO). Both methods have their respective uses. GWR is a regression based on the aspect of geogaphic location because heterogeniety test not fulfilled. LASSO is a regresi method for overcoming multicollinearity in data. From these two problems simultaneously in one regression model then the GWL method was created. In this study, we will apply and interpret the result of the GWL method in the Gross Regional Domestic Product (GRDP) problem in West Java. Based on the analysis results, 27 model were obtained with the R-Square value of the GWL method 0,9472643. One model is GRDP model for the city of bandung with an x_3value of 0,292049, with an x_8 value of 0,640292 and with an x_12 value of -0,01114. Keywords : Spasial Heterogeniety, Multicollinearity, Geographically Weighted Regression (GWR), Least Absolute Shringkage and Selection Operator (LASSO), Geographically Weighted Lasso (GWL), Gross Regional Domestic Product

Item Description

http://repository.upi.edu/46946/5/S_MAT_1501829_Title.pdf
http://repository.upi.edu/46946/2/S_MAT_1501829_Chapter1.pdf
http://repository.upi.edu/46946/7/S_MAT_1501829_Chapter2.pdf
http://repository.upi.edu/46946/1/S_MAT_1501829_Chapter3.pdf
http://repository.upi.edu/46946/3/S_MAT_1501829_Chapter4.pdf
http://repository.upi.edu/46946/4/S_MAT_1501829_Chapter5.pdf
http://repository.upi.edu/46946/6/S_MAT_1501829_Appendix.pdf