Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/166521| Title: | Pemodelan Berbasis Wilayah Data Dimensi Tinggi pada Kasus Stunting di Indonesia Menggunakan Generalized LASSO Termodifikasi |
| Other Titles: | Region-Based Modeling of High-Dimensional Data on Stunting Cases in Indonesia Using a Modified Generalized LASSO |
| Authors: | Rahardiantoro, Septian Wijayanto, Hari Darajati, Aida |
| Issue Date: | 2025 |
| Publisher: | IPB University |
| Abstract: | Stunting masih menjadi permasalahan kesehatan di Indonesia, meskipun prevalensinya menurun hingga 6,1% di tahun 2024, beberapa provinsi masih memiliki tingkat prevalensi stunting sangat tinggi. Penelitian ini bertujuan untuk mengeksplorasi pola stunting di seluruh wilayah Indonesia, menelaah peforma model generalized LASSO dalam identifikasi dugaan koefisien wilayah yang berdekatan pada beberapa ukuran ketetanggaan, serta menjelaskan peubah-peubah yang paling berpengaruh terhadap stunting di Indonesia. Data bersumber dari Badan Pusat Statistik dan Kementerian Dalam Negeri tahun 2024, mencakup 38 provinsi dengan sepuluh peubah bebas. Analisis dilakukan menggunakan modifikasi elastic net pada generalized LASSO melalui penambahan komponen penalti khusus (D2) ke dalam regularisasi L2 untuk mengurangi dampak korelasi antar peubah bebas. Pemilihan parameter tuning optimal dilakukan menggunakan metode ALOCV. Model KNN (k = 2) dengan matriks D2 terpilih sebagai model terbaik berdasarkan kombinasi optimal dari derajat bebas, CV-error, dan nilai RMSE. Hasil analisis menunjukkan bahwa peubah gini ratio (Papua), unmet need pelayanan kesehatan (Sulawesi, Maluku, NTT), akses sanitasi layak (Sumatra, Kalimantan, Jawa bagian barat), dan ASI eksklusif (Sumatra, Jawa bagian barat) merupakan faktor paling berpengaruh terhadap prevalensi stunting di Indonesia tahun 2024. Dalam penelitian ini, penambahan matriks D2 yang diusulkan mampu mengecilkan CV-error melalui identifikasi dampak serupa dari pasangan peubah bebas berkorelasi tinggi terhadap stunting. Stunting remains a health concern in Indonesia despite its prevalence decreasing to 6.1% in 2024, several provinces still have a very high prevalence of stunting. This research aims to explore the spatial patterns of stunting across Indonesia, examine the performance of the generalized LASSO model in identifying estimated adjacent region coefficients on several neighborhood measures, and explain the most influential variables on stunting prevalence in Indonesia. The data used is sourced from the Statistics Indonesia (BPS) and the Ministry of Home Affairs in 2024, covering 38 provinces with ten independent variables. The analysis was conducted using an elastic net modification on generalized LASSO by adding a specific penalty component (D2) to the L2 regularization in order to reduce the effect of correlation among independent variables. Optimal tuning parameters were selected using the ALOCV method. The K-Nearest Neighbors (KNN) model with k = 2 and the D2 matrix was selected as the best model based on an optimal combination of degrees of freedom, lowest CV-error, and RMSE value. The analysis identified the gini ratio (Papua), unmet need for healthcare services (Sulawesi, Maluku, East Nusa Tenggara), access to proper sanitation (Sumatra, Kalimantan, Jakarta, Banten, West Java), and exclusive breastfeeding (Sumatra, Jakarta, Banten, West Java) as the most influential variables on stunting prevalence in 2024. In this study, the addition of the proposed D2 matrix was able to reduce the CV-error by identifying similar effects from pairs of highly correlated predictor variables on stunting. |
| URI: | http://repository.ipb.ac.id/handle/123456789/166521 |
| Appears in Collections: | UT - Statistics and Data Sciences |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_G1401211016_c6d3f6358e824b9695616b5b53f95795.pdf | Cover | 547.17 kB | Adobe PDF | View/Open |
| fulltext_G1401211016_d9af4876d7fc4d778249d8596a2522e4.pdf Restricted Access | Fulltext | 850.51 kB | Adobe PDF | View/Open |
| lampiran_G1401211016_7ee5f51986824468a50188f8683055bc.pdf Restricted Access | Lampiran | 324.02 kB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.