COMPARISON OF THERMAL COMFORT PARAMETERS USING DIFFERENT FEATURE SELECTION ALGORITHMS

Authors : Resul Özlük, Türkan Göksal Özbalta

Pages : 1964-1974

Doi:10.31796/ogummf.1717391

View : 46 | Download : 125

Publication Date : 2025-12-19

Article Type : Research Paper

Abstract :This study aims to examine the environmental parameters affecting user thermal comfort in an amphitheater belonging to an educational building and to compare the effects of these parameters using different feature selection algorithms. Thermal comfort indicators PPD and environmental parameters were calculated using measurements from the Testo 480 device. Subjective data were obtained through surveys measuring users\\\' thermal acceptability and thermal comfort perception. The RandomForestRegressor, SHAP, and CorrelationAttributeEval algorithms were used to compare the order of influence of environmental parameters affecting students\\\' thermal comfort. The results showed that surface temperature and indoor air temperature are the most influential parameters on user comfort. In the SHAP and CorrelationAttributeEval algorithms, surface temperature was identified as the most influential parameter, while in the RandomForestRegressor algorithm, indoor temperature was identified as the most significant parameter. Additionally, when compared with survey results, the environmental parameter with the least effect on user comfort was found to yield the same result as the CorrelationAttributeEval algorithm. These findings provide important insights into better interpreting the factors affecting user comfort and optimizing thermal comfort in similar spaces.
Keywords : Isıl konfor, Yapay zeka algoritmaları, Eğitim binası, Çevresel parametreler, Özellik seçimi algoritmaları

ORIGINAL ARTICLE URL