Occupancy Prediction in Multi-Occupant IoT Environments Leveraging Federated Learning

Energy efficiency and energy saving have become crucial issues in the face of increasing energy demand, the need for sustainable solutions, and concerns about climate change. Buildings, as significant contributors to energy consumption and greenhouse gas emissions, require effective measures for energy optimization that can also be reached by predicting the usage of the building spaces. This paper introduces a data-driven approach combining Internet of Things sensors, Machine Learning, Edge computing, and Federated Learning to predict multi-occupancy in buildings. The proposed approach is used on real data from the ICAR-CNR IoT Laboratory in order to extract insights into occupancy patterns within a multi-occupant environment. Finally, a comparative analysis conducted by varying Federated Learning configurations demonstrates the robustness of the solution.