Recognition and prediction of individual thermal comfort requirement based on local skin temperature

This study focuses on identifying individual thermal comfort requirement using infrared images. The primary objective is to verify the feasibility of infrared images in predicting thermal comfort parameters, and to propose corresponding prediction models for two different classification modes defined in this study. The relevant data for modeling was obtained via chamber experiments. In addition, multiple machine learning (ML) algorithms, such as logistic regression (LR), support vector machines (SVM), and random forest (RF), were applied to propose models in which the local skin temperature (forehead, eye, nose, cheek and ear) measured using the infrared camera is used as an input parameter and thermal sensation vote (TSV) is used as an output parameter. The area under curve (AUC) of each model on the test set was calculated and used as the main evaluation index. In order to compare with the traditional thermal comfort model, the accuracy was also concerned as an evaluation index. The result shows that the prediction accuracy of ML models based on local skin temperature is close to that of predicted mean vote (PMV) model and indicates significant feasibility for practical application. Among the models, the AUC and accuracy of LR model reflects a relative advantage in comparison with other ML models and PMV model. The selection of measurement sites is also discussed in this study, and suggestions for practical application are given according to the performance of the models based on selected algorithms.

Automated summary

This study focuses on identifying individual thermal comfort requirements using infrared images and proposes prediction models based on machine learning algorithms. The results show that the machine learning models based on local skin temperature have similar prediction accuracy to traditional models and have significant feasibility for practical application.