Abstract
People usually spend several hours per day inside buildings, and they require great amounts of energy and resources to operate. Although there are numerous studies about smart buildings, there is still a need for new intelligent techniques for efficient smart building management. This paper proposes the use of Wi-Fi network association information as a basis for the design of intelligent systems for smart buildings. We propose a unified experimental methodology to evaluate machine learning (ML) models on their capacity to accurately predict Wi-Fi access point demand for energy-efficient smart buildings. The evaluation involves the use of multiple classification and regression models using a variety of configurations and algorithms. We conducted an experimental analysis using our proposed methodology to determine which ML models provide the best performance results using data collected from a large scale Wi-Fi network located at Fluminense Federal University (UFF) over a period of 6 months. The proposed methodology enables the user to evaluate and to create ML models for energy efficient smart building management systems. We achieved 86.69% accuracy for occupancy prediction using classification techniques and RMSPE (Root Mean Squared Percentage Error) of 0.29 for occupancy count prediction using regression techniques.
Highlights
Buildings play an important role in our lives
We show how our proposed unified methodology can help to select prediction models using machine learning for occupancy prediction based on performance metrics evaluation for distinct scenarios
We evaluated several regression models using SL and ML machine learning methods. 48 models were built using a combination of four distinct parameters: the SL method and 2 distinct ML (BR and RC) methods; 2 distinct types of model construction, which can be collective (Col) or individual (Ind); 2 distinct input configurations, one composed by APHDWD features and other by all features (ALL); and 4 distinct machine learning algorithms (RF, Decision Tree (DT), K-NN, XG) for constructing both SL models and the base classifiers of the ML methods
Summary
GUILHERME HENRIQUE APOSTOLO 1,2, FLAVIA BERNARDINI2, LUIZ C.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.