An important instrument for achieving smart and high-performance buildings is Machine Learning (ML). A lot of research has been done in exploring the ML models for various applications in the built environment such as occupancy prediction. Nevertheless, the research focused mostly on analyzing the feasibility and performance of different supervised ML models but has rarely focused on practical applications and the scalability of those models. In this study, a transfer learning method is proposed as a solution to typical problems in the practical application of ML in buildings. Such problems are scaling a model to a different building, collecting ground truth data necessary for training the supervised model, and assuring the model is robust when conditions change. The practical application examined in this work is a deep learning model used for predicting room occupancy using indoor climate IoT sensors. This work proved that it is possible to significantly reduce the length of ground truth data collection to only two days. The robustness of the transferred model was tested as well, where performance stayed on a similar level if a suitable normalization technique was used. In addition, the proposed methodology was tested with room occupancy level prediction, showing slightly lower performance. Finally, the importance of understanding the performance metrics is crucial for market adoption of ML-based solutions in the built environment. Therefore, in this study, additional analysis was done by presenting the occupancy prediction model performance in understandable ways from the practical perspective.

Climate change and technological development are pushing buildings to become more sophisticated. The installation of modern building automation systems, smart meters, and IoT devices is increasing the amount of available building operational data. The common term for this kind of building is a smart building but producing large amounts of raw data does not automatically offer intelligence that would offer new insights to the building’s operation. Smart meters are mainly used only for tracking the energy or water consumption in the building. On the other hand, building occupancy is usually not monitored in the building at all, even though it is one of the main influencing factors of consumption and indoor climate parameters. This paper is bringing the true smart building closer to practice by using machine learning methods with sub-metered electricity and water consumptions to predict the building occupancy. In the first approach, the number of occupants was predicted in an office floor using a supervised data mining method Random Forest. The model performed the best with the use of all predictors available, while from individual predictors, the sub-metered electricity used for office equipment showed the best performance. Since the supervised approach requires the continuous long-term collection of ground truth reference data (between one to three months, by this study), an unsupervised data mining method k-means clustering was tested in the second approach. With the unsupervised method, this study was able to predict the level of occupancy in a day as zero, medium, or high in a case study office floor using the equipment electricity consumption.