Forecasting Malaysia's energy consumption dynamics: A sectoral analysis using hybrid RF-ANN modeling and shared socioeconomic pathways

Accurately forecasting sectoral energy demand is critical for designing effective decarbonization strategies in a growing economy. This study supports Malaysia's long-term energy planning by developing a Bayesian-optimized hybrid Random Forest and Artificial Neural Networks (RF-ANN) model integrated with Shared Socioeconomic Pathways (SSPs) to project sectoral energy demand up to 2050. The proposed framework captures nonlinear and sector-specific dynamics often overlooked in previous works, while quantifying uncertainty through Probability Density Functions (PDFs). Five key sectors, i.e., agriculture, transport, residential and commercial, industrial, and non-energy, were analyzed under alternative socioeconomic trajectories (SSP1–SSP5). Results revealed that under the SSP1 (Sustainability-Focused Pathway), Malaysia's sectoral energy consumption is projected to reduce by 25% by 2050 compared to 2022 due to efficiency gains and renewable energy adoption, whereas the SSP5 (Fossil-Fueled) scenario is expected to rise by 78% in comparison to 2022 due to high economic growth and continued fossil fuel reliance, posing significant climate change risks. The integration of SSPs improves the interpretability of machine learning (ML)- based forecasting and provides scenario-driven insights for policymakers to evaluate risks, monitor progress, and adapt strategies within national policy and framework. This study demonstrates the potential of coupling advanced ML with SSP scenarios analysis to strengthen the evidence-based and sectoral energy transition planning and other developing economies.