Unsupervised Partial Domain Adaptation for Occupants Behavior Modeling in Smart Buildings

Jawher Dridi, Manar Amayri, Nizar Bouguila

2025

Abstract

Smart buildings rely on activity recognition (AR) and occupancy estimation (OE) tasks to provide residents with several services such as optimal energy management, HVAC (Heating, ventilation, and air conditioning) systems optimization, and security. Estimating the number of occupants and recognizing their activities is performed using sensor data which is scarce. The collection and labeling of smart building data are tedious, costly, and time-consuming, pushing researchers to consider solutions based on domain adaptation (DA) to transfer knowledge from source domains where data is abundant to target domains where data is scarce. In particular, unsupervised domain adaptation (UDA) has been considered to solve the unavailability of labeled data in target domains. Previous research has focused on standard UDA methods where label space is identical between source and target domains which is not the case for real-world datasets. This work considers unsupervised partial domain adaptation (UPDA) methods where target classes are a subset of source classes. We adapt and evaluate two UPDA techniques called Adversarial Re-weighting for Partial Domain Adaptation (ARPDA) and Selective Adversarial Networks for Partial Domain Adaptation (SAN w PDA). We have compared their performance to Adversarial Re-weighting for Standard Domain Adaptation (ARSDA) and Selective Adversarial Networks for Standard Domain Adaptation (SAN w SDA) as well as several previous UDA methods. The impressive results with scores up to 98% prove the efficiency of the adapted UPDA techniques. We provide the code in the following repository: https://github.com/JawDri/UPDA-for-OE-and-AR.git.

Paper Citation

in Harvard Style

Dridi J., Amayri M. and Bouguila N. (2025). Unsupervised Partial Domain Adaptation for Occupants Behavior Modeling in Smart Buildings. In Proceedings of the 14th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS; ISBN 978-989-758-751-1, SciTePress, pages 69-76. DOI: 10.5220/0013073400003953

in Bibtex Style

@conference{smartgreens25,
author={Jawher Dridi and Manar Amayri and Nizar Bouguila},
title={Unsupervised Partial Domain Adaptation for Occupants Behavior Modeling in Smart Buildings},
booktitle={Proceedings of the 14th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS},
year={2025},
pages={69-76},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013073400003953},
isbn={978-989-758-751-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 14th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS
TI - Unsupervised Partial Domain Adaptation for Occupants Behavior Modeling in Smart Buildings
SN - 978-989-758-751-1
AU - Dridi J.
AU - Amayri M.
AU - Bouguila N.
PY - 2025
SP - 69
EP - 76
DO - 10.5220/0013073400003953
PB - SciTePress