Nonlinear Modeling of Agricultural and Environmental SDG Indicators in ASEAN Using Extreme Learning Machine Algorithms

This study presents a data-driven approach for modeling Sustainable Development Goal (SDG) indicators in ASEAN countries using the Extreme Learning Machine (ELM) algorithm. Focusing on SDG 2 (Zero Hunger), SDG 6 (Clean Water and Sanitation), and SDG 15 (Life on Land), we utilized FAOSTAT datasets from 2020 to 2024 to forecast key indicators such as undernourishment, water use efficiency, and forest area. ELM, known for its rapid learning speed and capability to model nonlinear relationships, outperformed baseline models Linear Regression and Support Vector Machine (SVM) in terms of R² score, RMSE, and MAE. Specifically, ELM achieved R² values exceeding 0.93, with up to 54% RMSE reduction compared to linear models. The model successfully captured national development trends, including deforestation in Indonesia and Cambodia, water stability in Brunei, and varied progress in sustainable agriculture across the region. This study underscores the effectiveness of the Extreme Learning Machine (ELM) in forecasting Sustainable Development Goal (SDG) indicators and provides actionable insights to support evidence based policy planning, particularly in resource-constrained settings. The findings demonstrate that ELM’s combination of interpretability, computational efficiency, and scalability positions it as a highly valuable tool for real-time monitoring of sustainable development across Southeast Asia.