Personalized Temperature Settings: Machine Learning’s Contribution to Comfort


Regulating temperature in large buildings is crucial for ensuring occupants’ comfort and maximizing energy efficiency. However, traditional HVAC systems often struggle to strike a balance between these needs. Machine learning (ML) models offer a promising approach to predict occupants’ thermal perceptions and optimize building temperature control. Yet, these models may encounter challenges such as biased human-perception data and uncertainties, leading to inaccurate predictions and inefficient energy usage.

To address these issues, a team of researchers from Carnegie Mellon University’s Civil and Environmental Engineering Department proposed a novel method that combines data and models using Multidimensional Association Rule Mining (M-ARM) to identify and correct biases in human responses to temperature. Featured in Building and Environment, their research demonstrates that this method enhances the accuracy of predicting occupants’ comfort levels across various ML models.

The study leverages conflicting responses from building occupants to multiple thermal comfort questions to uncover the real “comfort zone” for the majority. By analyzing miscalibration issues and potential subjective biases in the data, M-ARM improves prediction reliability and reduces errors in current ML models.

Associate Professor Pingbo Tang, leading the study, emphasizes the potential of this approach to reduce energy consumption in large buildings while ensuring occupants’ comfort. Tang highlights the importance of considering factors beyond temperature, such as humidity and clothing, in assessing human comfort.

The research underscores the significance of leveraging question-answering behavior to adjust self-conflicts and estimate real thermal comfort. By considering various impact factors and calibration methods, the study demonstrates significant improvements in prediction reliability and model accuracy.

The findings offer valuable insights into advancing ML-based strategies for more reliable thermal perception predictions. Ultimately, this research could lead to more effective temperature control strategies in buildings, enhancing occupants’ comfort and reducing energy consumption.