Prediction of human thermal comfort preference based on supervised learning

Human thermal comfort is relevant to human life comfort and plays a pivotal role in occupational health and thermal safety. To ensure that intelligent temperature-controlled equipment can deliver a sense of cosiness to people while improving its energy efficiency, we designed a smart decision-making system that sets the thermal comfort adjustment preference as a label, reflecting both the human body's thermal feeling and its acceptance of the thermal environment. By training a series of supervised learning models underpinned by environmental and human features, the most appropriate adjustment mode in the current environment was predicted. To bring this design into reality, we tried six supervised learning models, and then, by comparison and evaluation, we identified that the Deep Forest's performance was the best. The model takes into account objective environmental factors and human body parameters. In this way, it can achieve high accuracy in application and good simulation and prediction results. The results can provide feasible references for feature selection and model selection in further research with the aim of testing thermal comfort adjustment preference. The model can provide recommendations for the thermal comfort preference in a specific place at a particular time, as well as guidance on human thermal comfort preference and thermal safety precautions in specific occupational groups.