# Are You Comfortable Now: Deep Learning the Temporal Variation in Thermal Comfort in Winters

**Authors:** Betty Lala, Srikant Manas Kala, Anmol Rastogi, Kunal Dahiya, Aya Hagishima  
**arXiv:** [2208.09628](https://arxiv.org/abs/2208.09628)  
**Date:** 2022-08-20  
**Summary page:** https://www.summarizepaper.com/en/paper/2208.09628v1/

## Key points

- Study focuses on indoor thermal comfort in smart buildings and its impact on occupants' health and performance
- Increasing use of machine learning (ML) to address challenges related to indoor thermal comfort, specifically temporal variability
- Importance of considering circadian rhythm and outdoor temperature in ML models for accurate predictions
- Data collected from month-long field experiments in 14 classrooms involving 512 primary school students
- Four thermal comfort metrics used as outputs for Deep Neural Networks and Support Vector Machine models
- Temporal factors such as time of day and outdoor temperature variations significantly influence prediction accuracy (up to 80% improvement)
- Contrasting micro-level (location-specific) and macro-level (city-wide) performance metrics highlight importance of spatio-temporal context
- Increase in time of day and sky illuminance leads to enhanced prediction accuracy for multiple thermal comfort metrics

## Summary

The study titled "Are You Comfortable Now: Deep Learning the Temporal Variation in Thermal Comfort in Winters" by Betty Lala, Srikant Manas Kala, Anmol Rastogi, Kunal Dahiya, and Aya Hagishima delves into the crucial role of indoor thermal comfort in smart buildings on the health and performance of occupants. The authors highlight the increasing use of machine learning (ML) to address challenges related to indoor thermal comfort, with a specific focus on the temporal variability of thermal comfort perception. In most ML-based studies on thermal comfort, temporal aspects such as the time of day, circadian rhythm, and outdoor temperature are often overlooked. This research aims to fill this gap by investigating how circadian rhythm and outdoor temperature impact the prediction accuracy and classification performance of ML models. The data for this study was collected through month-long field experiments conducted in 14 classrooms across 5 schools, involving 512 primary school students. The study considers four thermal comfort metrics as outputs for Deep Neural Networks and Support Vector Machine models applied to the dataset. Through a detailed analysis of the "time of day," the authors demonstrate how temporal variability influences school children's comfort levels. They show a significant improvement in prediction accuracy (up to 80%) when considering temporal factors such as outdoor temperature variations over time. Furthermore, the importance of spatio-temporal context is highlighted by contrasting micro-level (location-specific) and macro-level (6 locations across a city) performance metrics. The findings reveal that an increase in the time of day and sky illuminance leads to a definitive enhancement in prediction accuracy for multiple thermal comfort metrics. Overall, this study sheds light on the intricate relationship between circadian rhythm, outdoor temperature fluctuations, and temporal variability in predicting indoor thermal comfort levels. By incorporating these factors into ML models, researchers can better understand and optimize indoor environments for enhanced occupant well-being and energy efficiency.