COMPARING MACHINE LEARNING ALGORITHMS TO DIFFERENTIATE STRESS LEVELS USING EEG AND FNIRS

Researchers have difficulties accurately and objectively diagnosing stress in patients. The feeling of stress can often lead to serious health consequences if not properly monitored and treated. Therefore, this study aims to compare the effectiveness of various machine learning algorithms at differentiating levels of stress experienced by test subjects. This study analyzed the stress levels experienced by test subjects using electroencephalogram (EEG) and functional near-infrared spectroscopy (fNIRS) while they played a computer game that tested their decision-making and logical processing. Test subjects’ heart rates and average scores across each trial were also monitored and recorded. The EEG and fNIRS data were imported, processed, tested, and verified using MATLAB and created various machine learning algorithms for comparison based on accuracy. The data indicated the test subjects experienced increasing stress levels as the difficulty of the computer game increased. The machine learning algorithms varied in the accuracy in differentiating the levels of stress each test subject experienced but were overall successful in determining the stress experienced by each test subject