Thermal Comfort Modeling

Developed 1D Convolutional Neural Networks (CNNs) for time series classification of thermal preference based on physiological signals, achieving over 82% accuracy, outperforming traditional ML models.

Designed and evaluated deep learning architectures (Fully Convolutional Neural Network, Multi Convolutional Neural Network, Multi Convolutional Deep Neural Network) to capture temporal dependencies, resulting in 20% higher accuracy than ML models (Random Forest, ExtraTrees, C5.0).