Prediction of human thermal comfort preference based on supervised learning

Area of Science:

• Building Science
• Artificial Intelligence
• Human Factors Engineering

Background:

• Human thermal comfort is crucial for well-being, occupational health, and thermal safety.
• Intelligent temperature-controlled systems require accurate prediction of user comfort and environmental acceptance.
• Optimizing energy efficiency while maintaining comfort is a key challenge in building management.

Purpose of the Study:

• To design a smart decision-making system for predicting human thermal comfort adjustment preferences.
• To develop a model that balances user comfort with energy efficiency in intelligent environments.
• To identify the most effective machine learning model for predicting thermal comfort preferences.

Main Methods:

• Utilized supervised learning models trained on environmental and human physiological features.
• Evaluated six different supervised learning models to determine the best performing algorithm.
• Developed a system that labels thermal comfort adjustment preferences based on human feeling and environmental acceptance.

Main Results:

• The Deep Forest model demonstrated superior performance compared to other supervised learning models.
• The developed model accurately predicts the most appropriate thermal adjustment mode.
• Achieved high accuracy in application with strong simulation and prediction capabilities.

Conclusions:

• The Deep Forest model offers a reliable approach for predicting human thermal comfort preferences.
• Findings provide references for feature and model selection in thermal comfort research.
• The system can guide thermal comfort and safety recommendations for specific environments and occupational groups.