Evaluation and prediction of elderly thermal comfort at varying ambient temperatures based on electroencephalogram signals and machine learning

To explore the effects of ambient temperature on individual electroencephalogram (EEG) signals and predict thermal comfort in the elderly, continuous EEG signal recording of five elderly is conducted during the experimental procedure at four different temperature conditions (10/15/20/25°C). One-way repeated measure analysis of variance (ANOVA) is used to analyze the preprocessed EEG data and five identified bands in three channels are extracted as feature variables for developing a machine-learning-based personal thermal discriminant model. Four models are evaluated and the Random Forest classification model performs best with the highest prediction accuracy (81.217%) and the highest F1-score (0.783) in this case. This study can be utilized to assist thermoregulators in improving indoor thermal quality by monitoring EEG signals and predicting individual thermal comfort in real-time in future work.