Estimating occupancy from measurements and knowledge with Bayesian Networks

Manar Amayri (1) , Quoc-Dung Ngo , Stéphane Ploix (1)

1 G-SCOP_GCSP - Gestion et Conduite des Systèmes de Production

A general approach is proposed to determine the occupancy in a room using sensor data and knowledge coming respectively from observation and questioning are determined. Means to estimate occupancy include motion detections, power consumption and and acoustic pressure rewarded by a microphone. The proposed approach is inspired from machine learning. It starts by determining the most useful measurements in calculating information gains. Then, a non supervised estimation algorithm is proposed: it relies on Bayesian Network algorithms to model a human behaviour with probabilistic cause-effect relations based on knowledge and questioning. In addition, knowledge has been extracted from supervised learning algorithm. Bayesian Network (BN) based approach has been applied to an office setting, with an average estimation error of 0.09 and an accuracy of 90%. This approach avoids the usage of a camera to determine the actual occupancy required for supervised learning.

Mots clés: office buildings; building performance; activi- ties recognition; data mining; machine leaning; Index Terms—human behavior

Domaines: Statistiques [stat]; Machine Learning [stat.ML]

Manar Amayri, Quoc-Dung Ngo, Stéphane Ploix. Estimating occupancy from measurements and knowledge with Bayesian Networks. 2016 International Conference on Computational Science and Computational Intelligence (CSCI), Dec 2016, Las Vegas, United States. ⟨10.1109/CSCI.2016.0102⟩. ⟨hal-01864924⟩