AI-Based Stress State Classification Using an Ensemble Model-Based SVM Classifier

Abstract

The EEG signal is an electrical flow between brain neurons, and it appears differently depending on the mental and physical state. In this paper, stress is classified by analyzing EEG signals based on artificial intelligence. In this paper, using the DEAP dataset the stress state and the non-stress state were separated and trained in an artificial intelligence algorithm. As input to the AI algorithm, statistical, Power Spectrum Density (PSD), and High Order Crossings (HOC) features is used. These features were classified by learning the ensemble-based SVM classifier for each subject. To compare the classification accuracy, we compared the results using the feature selection algorithm using GA. In the comparison of experimental results, the ensemble-based SVM classifier showed better accuracy, such as 71.76% accuracy for feature selection using PCA and 77.51% accuracy for the experiment using ensemble-based SVM classifier.