Indoor Thermal Comfort Prediction Model for Patients in Rehabilitation Wards

This paper aims to propose an artificial neural network (ANN) based personal thermal comfort prediction model for inpatients. The indoor thermal environment affects occupant’s physical and psychological health, so it is vital to maintain it within comfort levels in the healthcare environment. Predicted Mean Vote (PMV), as the most popular model, has a limitation in processing various complex parameters and reflecting the individual occupant’s preference in thermal comfort. Some scholars utilized the machine learning (ML) method in exploring personal thermal comfort prediction because of its strong self-study, high-speed computing, and complex problem-solving abilities. However, there was a lack of relevant studies in the healthcare environment due to data collection difficulties and pathology complexity. The present research developed an ANN-based personal thermal comfort prediction model for patients in the healthcare environment. Ten-week fieldwork was conducted in an inpatient room to collect real-world environmental data, personal related information and thermal comfort voting for the model establishment. Additionally, the spatial variables and healthcare-related parameters (personal health information and medical treatment) were represented, and their impact on the model performance was explored. It is found that considering spatial parameters in the ANN-based model development has significantly increased the prediction accuracies compared with the conventional models. In addition, personal healthcare-related parameters also had some effect on the accuracy of model prediction.