The Development of Personalized Visual Comfort Prediction Models Using Machine Learning Algorithms

This study extends the framework of personal thermal comfort models to visual comfort prediction using machine learning. In the course of the experiment, twelve participants experienced three thermal and five lighting conditions over five consecutive working days, each evaluating visual comfort 140 times. Survey responses and measured environmental and physiological data revealed individual differences in comfort ratings, supporting the need for personalized lighting. Prediction models based on environmental conditions achieved up to 82% accuracy, while those using physiological responses reached 68%. Despite limitations in the variety of lighting conditions and a low frequency of pupil size measurements, the findings highlight a promising approach for personalized lighting solutions integrating personal preferences and environmental dynamics.