<![CDATA[Imbalanced data can reduce a classification model's ability to identify poor households as the minority class. This study identifies variables associated with household poverty status in D.I. Yogyakarta Province and evaluates the performance of Rare Event Weighted Logistic Regression (RE-WLR) under imbalanced data. Data were obtained from the March 2024 National Socio-Economic Survey (SUSENAS), with households as the unit of analysis. RE-WLR was fitted using several class-ratio schemes and regularization parameters and compared with standard binary logistic regression. Health complaints, beneficiary status of Program Keluarga Harapan (PKH), and educational attainment were significantly associated with poverty status. At the classification thresholds used, RE-WLR identified the minority class, whereas standard binary logistic regression with the default threshold tended to classify observations as the majority class. These findings suggest that RE-WLR is an alternative for household poverty classification under imbalanced data.]]>
Copyrights © 2026
