Personal thermal comfort models: a deep learning approach for predicting older people’s thermal preference
Smart and Sustainable Built Environment
ISSN: 2046-6099
Article publication date: 1 March 2022
Issue publication date: 5 July 2022
Abstract
Purpose
This paper presents the development of personal thermal comfort models for older adults and assesses the models’ performance compared to aggregate approaches. This is necessary as individual thermal preferences can vary widely between older adults, and the use of aggregate thermal comfort models can result in thermal dissatisfaction for a significant number of older occupants. Personalised thermal comfort models hold the promise of a more targeted and accurate approach.
Design/methodology/approach
Twenty-eight personal comfort models have been developed, using deep learning and environmental and personal parameters. The data were collected through a nine-month monitoring study of people aged 65 and over in South Australia, who lived independently. Modelling comprised dataset balancing and normalisation, followed by model tuning to test and select the best hyperparameters’ sets. Finally, models were evaluated with an unseen dataset. Accuracy, Cohen’s Kappa Coefficient and Area Under the Receiver Operating Characteristic Curve (AUC) were used to measure models’ performance.
Findings
On average, the individualised models present an accuracy of 74%, a Cohen’s Kappa Coefficient of 0.61 and an AUC of 0.83, representing a significant improvement in predictive performance when compared to similar studies and the “Converted” Predicted Mean Vote (PMVc) model.
Originality/value
While current literature on personal comfort models have focussed solely on younger adults and offices, this study explored a methodology for older people and their dwellings. Additionally, it introduced health perception as a predictor of thermal preference – a variable often overseen by architectural sciences and building engineering. The study also provided insights on the use of deep learning for future studies.
Keywords
Acknowledgements
The authors thank the participants involved in the research. The research has been funded by the Australian Research Council (Discovery Project number ARC DP180102019). Larissa Arakawa Martins is a recipient of the Faculty of the Professions Divisional Scholarship from The University of Adelaide and the Australian Housing and Urban Research Institute Supplementary Top-up Scholarship. The project has approval from The University of Adelaide Human Research Ethics Committee (approval number H-2018-042).
Declaration of interest statement: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Citation
Arakawa Martins, L., Soebarto, V., Williamson, T. and Pisaniello, D. (2022), "Personal thermal comfort models: a deep learning approach for predicting older people’s thermal preference", Smart and Sustainable Built Environment, Vol. 11 No. 2, pp. 245-270. https://doi.org/10.1108/SASBE-08-2021-0144
Publisher
:Emerald Publishing Limited
Copyright © 2022, Emerald Publishing Limited