Building Occupancy Detection using Machine Learning-based Approaches: Evaluation and Comparison

Building occupants and systems have huge impact on total energy use of the building due to their interaction with the building envelope and systems. The human factor and their behavior play a significant role in this process, and their influence must be considered. As a result, occupant behavior analysis and their detection, estimation, and prediction lead to an optimal energy and comfort management of buildings. Many methods for occupancy detection are presented in the literature, primarily machine learning methods due to their high performance accuracy and relatively simple design. However, there is no standard method for selecting algorithms on the basis of dataset type requirement (variables, number of occupants, etc.,) and list of better performing algorithms suitable for occupancy detection. In this paper, we have simulated and compared majorly used machine learning methods for occupancy detection using the open source Python platform. We evaluated performance in terms of accuracy, F1 score, area under the curve, and computational cost. Finally, some critical discussion and recommendations are proposed.