Abstract
Predicting thermal comfort in indoor environments is important for improving residents’ well-being, productivity, and energy efficiency. This study explores machine learning approaches, specifically Support Vector Machines (SVM) and Random Forest (RF), to improve thermal comfort prediction. Traditional methods rely on subjective assessments, whereas our approach leverages data-driven models trained on large thermal comfort datasets. The dataset underwent rigorous preprocessing, with 80% used for training and 20% for testing. The integration of the Internet of Things (IoT) further enhances predictive accuracy by enabling adaptive control in smart building systems. A comparative analysis of SVM and RF reveals that while both models effectively capture the complex interactions between environmental parameters and resident comfort, RF demonstrates greater stability and higher accuracy in most scenarios. The paper proposes potential strategies for integrating additional predictive features to further enhance model accuracy, demonstrating the advancement of machine learning in optimizing indoor comfort.