<![CDATA[Extreme poverty is the most severe form of poverty, characterized by a household's inability to meet basic needs and tends to persist despite ongoing social program interventions. In Jambi Province, poverty trends are fluctuating and influenced by macroeconomic dynamics and the agricultural sector; while extreme poverty indicators show an aggregate decline, inequality remains between districts/cities. This study aims to: (1) analyze socioeconomic factors influencing the extreme poverty status of households in Jambi Province, (2) compare the performance of prediction models using econometric approaches (logistic regression) and machine learning (Random Forest and XGBoost), and (3) examine differences in the determinants of extreme poverty between agricultural and non-agricultural households. The data used are SUSENAS microdata for the 2020–2024 period using a pooling approach (cross-section and time series) for all districts/cities in Jambi Province. Extreme poverty status is defined based on the international threshold of USD 2.15 PPP or national adjustment (TNP2K) in the relevant year. Modeling was performed by dividing the training data into 80% and 20% test data, conducting feature selection, model training, and hyperparameter tuning, as well as evaluation based on the confusion matrix and AUC–ROC. In addition to performance evaluation, this study emphasized sectoral comparative analysis by training the model separately on agricultural and non-agricultural subsamples to identify dominant determinants that are both universal and sector-specific.]]>
Copyrights © 2025
