Exploring the main and interaction effects of key factors on bedroom thermal perception using an interpretable machine learning approach

To create a comfortable and healthy indoor sleep environment and explore precise environmental control strategy, it is essential to understand the main and interaction effects between thermal perception and critical factors. However, the pattern of interactions between these factors remains unclear, especially for bedroom thermal perception. This investigation screened related records about bedrooms from the Chinese Thermal Comfort Dataset and used the cleaned data to train a RuleFit model. The H-statistic and Accumulated Local Effects Plot (ALE) were employed to interpret the machine learning model. Furthermore, the integrated impact of five key factors—air temperature, relative humidity, air velocity, clothing insulation, and metabolic rate—on subjective thermal perception was analyzed. The findings ranked the main effects and interaction effects on thermal perception in bedrooms and explored the relationship between main effects and interactions, although their importance varies among different conditions. The effect of two-way interactions is higher than that of three-way interactions, with an average intensity difference of more than five times. Furthermore, the influence of pure interactions on thermal perception may be positive or negative, depending on the levels of the interacting factors and their combination. And it should be emphasized that pure interactions are not dominated by main effects. Gender differences are also observed in the interaction effects. Compared to females, males are more susceptible to interaction effects, especially the interaction between air velocity and clothing insulation. The findings suggest that interaction effects, particularly two-way interactions, should be properly considered when designing bedroom thermal environments.