Application of Machine Learning in Occupant and Indoor Environment Behavior Modeling: Sensors, Methods, and Algorithms

The present chapter is focused on providing a comprehensive perspective of the applications of sensor-driven machine learning-based methodologies for occupant and indoor environment behavior modeling. In the first part of the chapter, various methodologies employed for non-intrusive occupancy status estimation, including the utilized sensors, feature generation methods, and detection algorithms, are reviewed. The second part is instead dedicated to comparing different methods that have been proposed in the literature for estimating the status of windows. Next, a thorough review on data-driven approaches utilized for simulating and predicting the thermal behavior of indoor environments is provided. Finally, the results of studies dedicated to machine learning-based occupancy prediction and implementing occupant-centered HVAC control are reviewed. For each case, the most promising set of sensors and algorithms, utilizing which has been proved in the previous studies to result in achieving a promising performance, have been provided. In addition, the methodologies that can be employed in order to simplify the corresponding pipelines, enhance the achieved accuracy, and facilitate the physical interpretation of the obtained results have also been discussed.