A Modified Single-Stacked Ensemble Regression Framework for Data-Driven Rice Yield Forecasting in Southeast Asia

Abstract

Food insecurity in Southeast Asia has intensified in recent years, highlighting the urgent need for reliable and interpretable quantitative methods for agricultural yield forecasting under data-limited conditions. This study develops a modified single-stacked ensemble statistical framework that integrates Simple Linear Regression (SLR) and Multiple Linear Regression (MLR) to enhance predictive robustness and interpretability. The proposed framework is implemented within the Cross Industry Standard Process for Data Mining (CRISP-DM), ensuring systematic data preparation, algorithm construction, and validation across five Southeast Asian nations, utilizing 28 potential determinants. From a numerical statistics perspective, the ensemble configuration is designed to improve predictive stability and reduce variance while maintaining algorithm transparency under constrained sample conditions. Comparative evaluation against selected conventional Artificial Intelligence (AI) algorithms demonstrates that the proposed regression-based ensemble achieves competitive and, in several cases, superior forecasting performance in terms of accuracy, generalizability, and computational efficiency. The results further quantify the interconnected effects of geographical characteristics, climate variables, and macro-structural factors on rice yield dynamics. By providing a scalable and computationally efficient modeling approach, this framework offers a practical foundation for applied mathematical and statistical applications in data-driven agricultural decision-making. The findings also offer actionable insights to support evidence-based policymaking and related strategic investment, contributing indirectly to the advancement of regional food security and related Sustainable Development Goals (SDGs), including SDG1 (No Poverty), SDG2 (Zero Hunger), SDG3 (Good Health and Well-being), and SDG8 (Decent Work and Economic Growth).