Developing the Reliable Shallow Supervised Learning for Thermal Comfort using ASHRAE RP-884 and ASHRAE Global Thermal Comfort Database II
This work addresses data reliability issues for AI designers in thermal comfort, particularly for IoT systems in residential use, but it is incremental as it builds on existing datasets and methods.
The paper tackles the problem of insufficient or unreliable data for training AI systems in thermal comfort, introducing a control algorithm based on shallow supervised learning that uses ASHRAE databases, with methods for data filtering and augmentation to address overfitting.
The artificial intelligence (AI) system designer for thermal comfort faces insufficient data recorded from the current user or overfitting due to unreliable training data. This work introduces the reliable data set for training the AI subsystem for thermal comfort. This paper presents the control algorithm based on shallow supervised learning, which is simple enough to be implemented in the Internet of Things (IoT) system for residential usage using ASHRAE RP-884 and ASHRAE Global Thermal Comfort Database II. No training data for thermal comfort is available as reliable as this dataset, but the direct use of this data can lead to overfitting. This work offers the algorithm for data filtering and semantic data augmentation for the ASHRAE database for the supervised learning process. Overfitting always becomes a problem due to the psychological aspect involved in the thermal comfort decision. The method to check the AI system based on the psychrometric chart against overfitting is presented. This paper also assesses the most important parameters needed to achieve human thermal comfort. This method can support the development of reinforced learning for thermal comfort.