Description Thermal comfort models can predict human response to sensible and latent loads in the environment. These models have achieved certain degree of success on predicting human response, but accuracy is still far from ideal. This project explores means to enhance such indexes based on the analysis of a large experimental dataset containing boundary conditions and human responses. Key Objectives Compare existing empirical datasets with state-of-the-art transient thermal comfort models. Analyse datasets using clustering techniques and sensitivity analysis. Evaluate the performance of machine learning techniques to improve prediction of existing thermal comfort models.