<![CDATA[Indonesia is accelerating its energy transition to reduce reliance on fossil fuels and achieve a 23% renewable energy mix by 2025. Bioethanol has emerged as a promising renewable alternative, particularly under the current E5 (5% ethanol blend) program and the roadmap toward higher integration. This study presents a Machine Learning (ML) framework for predicting cassava-based bioethanol production in West Java, applying Random Forest (RF) modeling to estimate cassava yields and a conversion process to calculate ethanol output. The dataset comprises 3,000 monthly observations from 27 districts (2013–2024), integrating agronomic, climatic, and economic variables such as harvested area, rainfall, temperature, and farm-gate price. Model validation using blocked time-series Cross-Validation (CV) demonstrated high predictive accuracy ( ; Root Mean Square Error (RMSE) = 191 kL; Mean Absolute Error (MAE) ≈ 40 kL), confirming the effectiveness of ensemble learning in capturing nonlinear agricultural dynamics. Feature importance analysis identified harvested area and rainfall as the most influential predictors, underscoring the relevance of climate-adaptive farming and sustainable land-use strategies. Spatial evaluation highlighted Garut, Sukabumi, and Tasikmalaya as priority districts. Aggregated projections, recalculated with verified conversion factors (=5.86 megawatt-hour per kiloliter, MWh/kL), indicate an energy potential of 820,000 MWh/year and substantial Carbon Dioxide (CO₂) reduction under E5 blending. Despite limitations related to data inconsistencies in productivity units and fluctuating price records, the findings demonstrate that cassava-based bioethanol can serve as a viable pathway for renewable energy diversification, provided that agricultural productivity, infrastructure, and policy alignment are strengthened.]]>
Copyrights © 2026
