Automated classification of indoor environmental quality control using stacked ensembles based on electroencephalograms

Abstract This study aims to develop an automated classification of indoor air quality control using the machine learning based on an electroencephalogram (EEG) signal. Two experiments were conducted: The first aimed to select an EEG channel based on the brain wave indices when the indoor environmental quality (IEQ) changes. We then extract the brain wave indices’ feature when the subjects conduct adaptive behaviors and predict the IEQ condition control using machine learning including the stacked ensembles. The extracted features were classified using base estimators such as distributed random forest, gradient boosting machine (GBM), generalized linear models (GLMs), deep neural network, and new predicted data were retrained and predicted by metalearner (i.e., GLM). In Dimension 1, the air conditioning system and the air ventilation system, and the area under curve (AUC) of the proposed stacked ensembles trained by base estimators was the highest, 0.9038. In Dimension 2, turning on and turning off, the AUC of the GBM is the highest, 0.8384. Based on these results, the EEG signal can be used to suggest an automatic IEQ control model that can then reduce the drowsiness, and increase attention.