Pemodelan Geographically Neural Network Weighted Regression pada Tingkat Pengangguran Terbuka di Indonesia
DOI:
https://doi.org/10.26740/mathunesa.v14n1.p634-647Abstract
Abstrak
Tingkat Pengangguran Terbuka (TPT) di Indonesia masih menjadi isu strategis dalam pembangunan ekonomi karena mencerminkan ketidakseimbangan antara pertumbuhan angkatan kerja dan ketersediaan lapangan kerja. Ketimpangan wilayah serta perbedaan kondisi sosial ekonomi antar daerah menyebabkan variasi spasial dalam tingkat pengangguran yang belum sepenuhnya dapat dijelaskan oleh model regresi klasik. Penelitian ini bertujuan untuk memodelkan faktor-faktor yang memengaruhi TPT di 512 kabupaten/kota di Indonesia tahun 2024 serta membandingkan performa beberapa pendekatan spasial. Metode yang digunakan adalah Geographically Neural Network Weighted Regression (GNNWR). Hasil menunjukkan bahwa setiap wilayah memiliki nilai koefisien pengaruh yang berbeda, menandakan adanya heterogenitas spasial pengaruh variabel terhadap TPT. Pada model GNNWR, variabel Indeks Pembangunan Manusia (IPM), Rata-rata Lama Sekolah (RLS), Persentase Penduduk Miskin (PPM), dan Tingkat Partisipasi Angkatan Kerja (TPA) berpengaruh signifikan terhadap TPT. Model GNNWR juga memberikan performa terbaik dengan nilai sebesar 0,68 (training) dan 0,61 (testing), AICc terendah (876,33), serta MAPE terkecil (20,01%). Hasil ini menunjukkan bahwa integrasi pendekatan spasial dan neural network pada model GNNWR mampu menangkap hubungan nonlinier dan meningkatkan akurasi prediksi TPT antarwilayah.
Kata Kunci: Geographically Neural Network Weighted Regression, Heterogenitas Spasial, Tingkat Pengangguran Terbuka
Abstract
The Open Unemployment Rate (OUR) in Indonesia remains a strategic issue in economic development, as it reflects the imbalance between the growth of the labor force and the availability of job opportunities. Regional disparities and differences in socioeconomic conditions across areas lead to spatial variation in unemployment rates that cannot be fully explained by classical regression models. This study aims to model the factors influencing the OUR in 512 regencies/municipalities in Indonesia in 2024 and to compare the performance of several spatial approaches. The method employed is Geographically Neural Network Weighted Regression (GNNWR). The results show that each region has different coefficient values, indicating spatial heterogeneity in the effects of variables on the OUR. In the GNNWR model, the Human Development Index (HDI), Mean Years of Schooling (MYS), Percentage of Poor Population (PPP), and Labor Force Participation Rate (LFPR) have a significant effect on the OUR. The GNNWR model also provides the best performance, with an R² of 0.68 (training) and 0.61 (testing), the lowest AICc (876.33), and the smallest MAPE (20.01%). These results indicate that integrating spatial approaches and neural networks in the GNNWR model is able to capture nonlinear relationships and improve the accuracy of interregional OUR predictions.
Keywords: Geographically Neural Network Weighted Regression, Spatial Heterogeneity, Open Unemployment Rate.
Downloads
Downloads
Published
Issue
Section
License
Copyright (c) 2026 MATHunesa: Jurnal Ilmiah Matematika

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Abstract views: 0
,
PDF Downloads: 0









