Cambodia's rice sector has undergone significant transformation over the past decades, evolving into a cornerstone of the national economy. Despite improvements in cultivation techniques and land use, the sector remains vulnerable to climate variability, inefficient resource allocation, and limited data-driven planning. To address these challenges, this study presents a comprehensive prediction framework that leverages both traditional Multiple Linear Regression (MLR) and advanced tree-based ensemble learning methods such as Random Forest, Gradient Boosting, and XGBoost to predict Cambodia’s annual rice production based on rice and climatic data of Cambodia. A Cambodia comprehensive dataset, compiled from the Food and Agriculture Organization (FAO) and World Bank, was used to train and evaluate the models, incorporating variables such as harvested area, annual production record, precipitation, and surface temperature. The results reveal that XGBoost achieved the highest predictive accuracy (R² = 0.9511), while feature importance analysis highlights the dominant influence of harvested area on rice yield. This work marks one of the first applications of ensemble learning in Cambodia's agricultural predictions and offers a decision – support tool that aligns with the nation's Digital Economy and Industrial Development Policies. The findings also contribute to national strategies on food security, smart agriculture, and enhance the global Sustainable Development Goals (SDGs).