Mental Stress and Relaxation Conditions of Engineering Students based on EEG Brainwave Signals using Machine Learning: A Case Study

The rising prevalence of mental health challenges among engineering students highlights the necessity for improved methods in recognising and managing stress. Conventional self-evaluation techniques seem to be subject to interpretation when it comes to assessing stress levels. This research examines the identification of electroencephalography (EEG) patterns that could serve as markers for mental stress, along with the classification of stress and relaxation mental states using machine learning methods. Earlier studies applied machine learning methods, particularly clustering algorithms, to analyse EEG data; nonetheless, a more targeted approach is required for improved accuracy in stress detection. This study employs EEG data that has undergone processing through feature extraction methods to develop a machine learning model aimed at distinguishing between states of stress and relaxation in the brain. EEG signals were collected using OpenBCI software, and feature extraction was conducted employing Fast Fourier Transform (FFT). Subsequently, four distinct machine learning algorithms were employed to categorise the brainwave signals: K-Nearest Neighbour (KNN), Support Vector Machine (SVM), Wide Neural Network (WNN), and Naïve Bayes (NB). The WNN achieved an impressive accuracy rate of 90%, with a Receiver Operating Characteristic (ROC) area analysis revealing a value of 0.88, which suggests favourable outcomes in distinguishing between stress and relaxation states. The findings contribute to the domain of mental health research by offering a dependable method for distinguishing and categorising stressed versus relaxed students, possibly improving future evaluation instruments for mental well-being in engineering students.