Perbandingan Performa Metode Klasifikasi Regresi Logistik, Classification Tree, dan Random Forest (Studi kasus: Perkawinan Anak pada Perempuan Usia Muda di Nusa Tenggara Barat Tahun 2022)

Pangestika, Dhita Elsha
Mustika, Diva Arum
Rahman, Ayub Abdul

Prosiding Seminar Nasional Official Statistics
Vol 2025 No 1 (2025): Seminar Nasional Official Statistics 2025

Child marriage remains a persistent issue in Indonesia, particularly in West Nusa Tenggara Province. This study compares the performance of three classification methods—logistic regression, classification tree, and random forest—in predicting child marriage among young women. The analysis uses 2022 National Socio-Economic Survey (Susenas) data, which comprises 69 women aged 20–24 who had married and were still living with their parents. Model performance was evaluated using the Area Under the Curve (AUC) metric with 50 validation repetitions. Logistic regression yielded the highest AUC (77.86%), followed by random forest (76.07%) and classification tree (75.49%). These results indicate that logistic regression is more stable and suitable for linear, low-dimensional, and limited observational data. Additionally, education level and the household head’s type of employment were identified as key predictors of child marriage.