Machine Learning Approach for Stress Detection based on Alpha-Beta and Theta-Beta Ratios of EEG Signals

The contribution to stress detection and classification is far beyond demand as the statistics show that the health and mental illness of society have kept on deteriorating. Electroencephalogram (EEG) signals have the potential to detect stress levels reliably due to their high accuracy. Majority of studies of stress detection are based on alpha and beta waves and the corresponding ratio of the two waves and there are hardly any based-on theta waves. This work explores the impact of bandpower of alpha/beta and theta/beta ratios when combined with other features to classify two-levels of human stress based on EEG signals using five commonly used machine learning algorithms. A classification model is developed from the clustering model gained and Naïve Bayes shows the highest accuracy which is 95% in compared to the other four common machine learning algorithms (i.e., SVM, Logistic, IBk, and SGD) by using WEKA. The proposed framework recommends that both ratios are reliable features, and theta/beta appears to give a huge impact compared to alpha/beta. This study will ultimately contribute to society's development with improved robust machine learning algorithm for binary classification.