Forecasting of Poverty using the Ensemble Learning Classification Methods

Abstract

Poverty is a social-cultural problem that can be categorized into monetary approach, capability approach, social exclusion, and participatory poverty assessment. However, the existing measurement methods are complex, costly, and time-consuming. This research was conducted to forecast poverty using classification methods. Random Forest and Extreme Gradient Boosting (XGBoost) algorithms were applied to forecast poverty since they are supervised learning algorithms that use the ensemble learning approach for classification. Ensemble Learning has improved the classification of poverty and obtained better predictive performance. The results of the algorithms showed the poverty trend, which helped to determine the poverty classification. Hence, this method will help the government to act and produce a specific plan to reduce the poverty rate. It is a strategic move to reduce global poverty, parallel to Goal 1 of Sustainable Development Goal (SDG): No Poverty