Darmakusuma, Ketut Bagus Ananta (2026) Analisis Faktor-Faktor Yang Memengaruhi Kejadian Bayi Berat Lahir Rendah (BBLR) Menggunakan Metode Elastic Net Ordinal Logistic Regression (Studi Kasus: RSUD Haji Provinsi Jawa Timur). Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5003221131-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5003221131-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Berat Bayi Lahir Rendah (BBLR) merupakan indikator penting derajat kesehatan masyarakat yang memiliki risiko tinggi terhadap mortalitas neonatal. Di RSUD Haji Provinsi Jawa Timur, kejadian BBLR memiliki tingkatan kategori keparahan (BBLR, BBLSR, dan BBLER) yang secara analitik umumnya dimodelkan menggunakan Ordinal Logistic Regression (OLR). Namun, penggunaan banyak variabel prediktor kondisi maternal sering kali memunculkan masalah multikolinearitas yang mengakibatkan varians estimator menjadi bias dan model tidak presisi. Penelitian ini bertujuan untuk mengatasi ketidakstabilan tersebut dengan menerapkan metode Elastic Net Ordinal Logistic Regression (ENOLR). Metode ENOLR mengintegrasikan fungsi penalti gabungan LASSO dan Ridge untuk melakukan penyusutan koefisien (shrinkage) dan seleksi variabel secara simultan. Estimasi parameter diselesaikan menggunakan algoritma Coordinate Descent, dengan parameter tuning optimal yang dievaluasi berdasarkan kriteria Generalized Cross-Validation (GCV). Hasil pengujian menunjukkan bahwa model OLR standar terbukti gagal memenuhi uji kesesuaian model akibat tingginya korelasi antara variabel Kadar Hemoglobin dan Lingkar Lengan. Sebaliknya, penerapan ENOLR dengan parameter optimal λ_EN = 2,3719 dan ϑ_EN = 0,9000 berhasil mengeliminasi variabel yang tumpang tindih dan mempertahankan Usia Ibu, Paritas 2, serta Usia Kehamilan sebagai prediktor yang valid secara medis. Usia Kehamilan dan Paritas 2 teridentifikasi sebagai faktor paling dominan yang memengaruhi klasifikasi tingkat keparahan BBLR. Kinerja ENOLR terbukti lebih unggul, parsimoni, dan akurat dibandingkan OLR konvensional, yang ditunjukkan oleh penurunan nilai Akaike Information Criterion (AIC) dari 110,8561 menjadi 106,7925, serta peningkatan Akurasi Klasifikasi dari 78,43% menjadi 81,37%.
=======================================================================================================================================
Low Birth Weight (LBW) is a crucial indicator of public health status associated with a high risk of neonatal mortality. At RSUD Haji East Java Province, the incidence of LBW is classified into several severity categories (LBW, Very Low Birth Weight/VLBW, and Extremely Low Birth Weight/ELBW), which are analytically typically modeled using Ordinal Logistic Regression (OLR). However, the inclusion of numerous maternal condition predictor variables often leads to multicollinearity issues, causing biased estimator variances and imprecise models. This study aims to address this instability by applying the Elastic Net Ordinal Logistic Regression (ENOLR) method. The ENOLR method integrates the combined penalty functions of LASSO and Ridge to perform simultaneous coefficient shrinkage and variable selection. Parameter estimation was conducted using the Coordinate Descent algorithm, with optimal tuning parameters evaluated based on the Generalized Cross-Validation (GCV) criterion. The test results showed that the standard OLR model failed the goodness-of-fit test due to the high correlation between Hemoglobin Levels and Mid-Upper Arm Circumference (MUAC) variables. In contrast, the implementation of ENOLR with optimal parameters of λ_EN = 2,3719 and ϑ_EN = 0.9000 successfully eliminated overlapping variables and retained Maternal Age, Parity 2, and Gestational Age as medically valid predictors. Gestational Age and Parity 2 were identified as the most dominant factors influencing the classification of LBW severity levels. The performance of ENOLR proved to be more robust, parsimonious, and accurate than conventional OLR, as indicated by a decrease in the Akaike Information Criterion (AIC) value from 110.8561 to 106.7925, and an increase in Classification Accuracy from 78.43% to 81.37%.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | BBLR, Coordinate Descent, Elastic Net, Multikolinearitas, Regresi Logistik Ordinal, Coordinate Descent, Elastic Net, LBW, Multicollinearity, Ordinal Logistic Regression |
| Subjects: | Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA278.3 Structural equation modeling. Q Science > QA Mathematics > QA401 Mathematical models. |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Ketut Bagus Ananta Darmakusuma |
| Date Deposited: | 11 Aug 2026 02:51 |
| Last Modified: | 11 Aug 2026 02:51 |
| URI: | http://repository.its.ac.id/id/eprint/144222 |
Actions (login required)
 |
View Item |