Supervised machine learning for thermal comfort and energy efficiency: An evaluation for the indoor built environment

The growing demand for energy-efficient and sustainable buildings has accelerated the exploration of advanced technologies to optimize thermal comfort and reduce energy consumption. Machine learning techniques, particularly supervised learning approaches, have shown strong potential to optimize HVAC control while maintaining comfort. However, existing studies are often fragmented, with limited integrated analyses of methodologies and applications, particularly in the context of diverse climates, building typologies, and occupant behaviors. This study addresses these gaps through a semi-systematic review of peer-reviewed studies applying supervised machine learning techniques for thermal comfort prediction and energy optimization. Using a transparent process involving Web of Science search, predefined inclusion/exclusion criteria, and Rayyan-assisted screening, 18 supervised learning articles were identified from an initial 603 records. These articles were categorized into tree-based models, regression-based models and neural networks. The review identifies critical gaps, such as the insufficient integration of real-time occupant behavior, limited applicability across diverse climatic conditions, and challenges in achieving a balance between energy efficiency and occupant comfort. Findings highlight the strengths of tree-based models in feature selection and real-time decision-making, the simplicity of regression-based models for controlled environments, and the adaptability of neural networks in complex, non-linear scenarios. Despite these advancements, limitations such as data scarcity, computational demands, and the lack of long-term validation persist. Addressing these challenges is essential for the development of robust and scalable machine learning-driven solutions. This study provides a roadmap for future research and practical applications, emphasizing the transformative potential of supervised machine learning techniques in achieving sustainable, energy-efficient, and occupant-centered building environments.