PEMODELAN GEOGRAPHICALLY WEIGHTED REGRESSION MENGGUNAKAN PEMBOBOT KERNEL FIXED DAN ADAPTIVE PADA KASUS TINGKAT PENGANGGURAN TERBUKA DI INDONESIA

Tingkat Pengangguran Terbuka (TPT) merupakan indikator untuk meng�ukur angka pengangguran menurut konsep ketenagakerjaan. Jumlah TPT mengalami peningkatan dalam setahun terakhir sebesar 1.84% dimana Agus�tus 2019 berada di angka 5.23% dan Agustus 2020 menjadi 7.07%, hal demikian terjadi karena dampak dari pandemi Covid-19 di Indonesia. Salah satu ana�lisis untuk mengetahui faktor-faktor yang mempengaruhi TPT di Indonesia yaitu regresi linear berganda dengan metode Ordinary Least Square (OLS). Penelitian dengan metode OLS menghasilkan gejala heterokedastisitas, me�nandakan bahwa data memiliki informasi lebih. Pengamatan berlanjut pada pengecekan terdapatnya aspek spatial (lokasi). Hasil tipe data menunjukkan bahwa penelitian menggunakan data spasial karena mengandung heterogeni�tas spasial. Data spasial merupakan data yang mengandung infomasi lokasi (latitude,longitude) dan informasi deskriptif (attribute) . Analisis data spasial menggunakan pendekatan titik yaitu dengan metode Geographically Weighted Regression (GWR). Analisis metode GWR merupakan pengembangan regresi global menjadi regresi terboboti, sehingga menghasilkan model yang bersifat lokal. Penulis menggunakan GWR dengan kernel pembobot Fixed Gaussian, Adaptive Gaussian, Fixed Bi-Square, dan Adaptive Bi-Square. Perbanding�an pemodelan untuk kasus pengangguran di Indonesia antara Regresi Linear Berganda dengan model GWR menghasilkan bahwa GWR Adaptive Bi-Square lebih baik, meninjau dari nilai R2 ,AIC dan JKG. Kemampuan model GWR Adaptive Bi-Square menjelaskan pengaruh TPT terhadap faktor-faktor (Ang�katan kerja, Keluhan Kesehatan dan Persentase Kemiskinan) sebesar 89.1% sedangkan model regresi global sebesar 46.1%. Unemployment Rate (UR) is an indicator for measuring the unemployment according to the concept of employment. Number of UR increased in the last year by 1.84% where August 2019 was at 5.23% and August 2020 to 7.07%, this is due to the impact of the Covid-19 pandemic in Indonesia. One of the ana�lysis to find out the factors that affect TPT in Indonesia is by using multiple linear regression with the Ordinary Least Square (OLS) method. Research with the OLS method produces symptoms of heterokedasticity , indicating that Data has more information. Observation continues on checking the spatial aspects ( location ). Data type results show that research uses spatial data because it contains spatial heterogenity Research with OLS methods produces symptoms of heterocedasticity, indicating that Data has more information. Observation continues on checking the spatial aspects ( location ). Data type results show that research uses spatial data, because it contains spatial heterogenity. Data Spatial is data that contains location information (latitude, longitude) and de�scriptive information (attributes). Spatial data analysis using a point approach is by the Geographically Weighted Regression method (GWR). Analysis of the GWR method is the development of global regression into weighted regression, resulting in a model that is local. The author uses GWR with Fixed Gaussian, Adaptive Gaussian kernel weighting, Fixed Bi-Square, and Adaptive Bi-Square. Comparison of modelling for Unemployment cases in Indonesia between Multi�ple Linear Regression and GWR model produces that GWR Adaptive Bi-Square better, review value of the R2,AIC and JKG . The ability of the GWR Adaptive Bi-Square model explains the effect of UR on factors (Labor Force or econo�mically active, Health Complaint and Poverty Percentage) by 89.1% while the global regression model 46.1%