This critical review explores how Machine Learning (ML) can be leveraged to integrate acoustics and sound perceptions into indoor comfort assessments. The study performed an in-depth analysis of 11 papers to identify the most commonly used tools, their advantages, potential, and limitations. Results show that Random Forest (RF) approach, one of the ML algorithms combining multiple decision trees to improve accuracy, is the most widely adopted prediction method. Results further emphasize the research gap and scarcity of studies combining ML with physiological measurement in acoustic experiences, urging further exploration in this area and, more broadly, multiple sensory perceptions to enhance predictive analysis and improve occupant comfort and well-being.