Modeling thermal sensation in naturally ventilated Mediterranean offices using air-to-radiant temperature offset and machine learning
Architectural Engineering and Design Management, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Basım Tarihi: 2026
- Doi Numarası: 10.1080/17452007.2026.2647808
- Dergi Adı: Architectural Engineering and Design Management
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, Environment Index, Index Islamicus, INSPEC
- Anahtar Kelimeler: machine learning, Mediterranean climate, natural ventilation, office buildings, radiant temperature, Thermal comfort
- Süleyman Demirel Üniversitesi Adresli: Evet
Özet
This study investigates the role of the air-to-radiant temperature offset, defined as the difference between air temperature and mean radiant temperature, in predicting thermal sensation in naturally ventilated office buildings located in warm Mediterranean climates. Although machine learning has advanced thermal comfort modeling by capturing complex relationships among environmental and physiological variables, most models continue to treat radiant effects indirectly, blending them into operative temperature rather than modeling the offset as a distinct predictor. This approach limits their ability to reflect the directional influence of radiative asymmetries commonly found in naturally ventilated environments. To address this gap, a data-driven approach using machine learning was employed. A curated subset of the ASHRAE Global Thermal Comfort Database II was used, filtered for naturally ventilated office spaces in the Csa climate zone during warm seasons. Eleven regression algorithms were compared to predict occupants’ thermal sensation votes based on operative temperature, air-to-radiant temperature offset, relative humidity, air velocity, and metabolic rate. Among these, a tuned Random Forest model achieved the highest performance, significantly outperforming traditional thermal comfort predictions. Feature importance and SHAP analysis revealed that the air-to-radiant temperature offset was an influential predictor, with clear non-linear effects on perceived comfort.