Intelligent classification of multi-layer clothing for dynamic thermal comfort assessment in energy-efficient environments

The intelligent clothing classification method that can be adjusted to different thermal conditions is an essential measure to maintain optimal thermal comfort in energy-saving settings. The traditional models ignore differences in garment insulation made by layering. To solve this, it suggests the Intelligent Multi-layer Clothing Classification (IMLCC) model, which combines the Scalable Monarch Butterfly Optimizer-powered Attention-Based Variational Autoencoder (SMB-AVAE) to determine the dynamic thermal comfort. The model groups clothing layers according to the type of fabric, its thickness, shape and coverage in a way that maximizes predictions of heat transfer. The sensor and physiological thermal responses are measured by IoT-enabled sensors and wearable devices respectively. The accuracy of classification is improved by preprocessing techniques, such as noise reduction. Scalable Monarch Butterfly Optimization (SMBO) is an optimizer of model parameters, making it faster to converge. Attention-Based Variational Autoencoder (AVAE) is an encoder of clothing that captures the characteristics of the garment in a latent space and retains the details of the texture to be accurately classified. Experimental results show that SMB-AVAE achieves high classification accuracy, recognizing at least three garments per ensemble. The performance measures of the proposed method are MAE (0.071), MSE (0.009) and R2 (0.98).