HUMAN WORK-RELATED STRESS CLASSIFICATION USING A CONSUMER-GRADE ELECTROENCEPHALOGRAPHIC DEVICE AND MACHINE LEARNING MODELS

Work-related stress manifests itself among workers when the demands made on them exceed their ability to cope, with damaging consequences for their health and mental equilibrium. It is therefore essential to constantly monitor their pathophysiological mechanisms. Stress is a common experience in everyday life. In biological terms, it represents the set of physiological responses that have evolved over millennia to enable us to react quickly in emergency situations and to keep us safe. It is therefore a natural, normally positive, and useful phenomenon. However, the incessant action of stressful factors can undermine physical and mental health, so the aim of this project is to carry out a binary classification of mental workload (MWL) using a particular app able to train models to classify data and monitor stress. From a scientific point of view, however, it is something different: stress is defined as the reaction of an organism to a stressful agent, or stressor. In fact, therefore, it is possible to say that stress is an entirely natural and fundamental phenomenon for organisms, because it puts them in a position to react to a potentially dangerous situation. The brain is an organ divided into two hemispheres joined by the corpus callosum, which communicates with the spinal cord through the brainstem. Its outermost layer is the cortex, while at its center are the basal ganglia and at its base, posteriorly, the cerebellum. Each hemisphere is divided into several lobes: frontal, parietal, occipital and temporal. Two types of cells are sufficient to form all these structures: neurons and glia. The brain is surrounded by membranes, the meninges, which form a triple protective layer. The brain controls thoughts, memory and language, the movements of the arms and legs and the functioning of all the body's organs. Finally, by regulating breathing and heartbeat, it determines reactions to stressful events that may occur in daily life. The central nervous system plays a key role in processing stressors and eliciting and controlling the stress response. Indeed, all types of stressful stimuli converge on the central nervous system where they are processed, and stress responses are triggered. Recently, considerable progress has been made in discovering the nervous and neuroendocrine correlates that mediate the cascade of reactions triggered by stress. 3 The non-invasive recording of electrical activity in the brain using external electrodes is called an electroencephalogram (EEG). The electrical activity detected is represented by a series of waves reproduced on a screen and then printed on paper or transferred to an electronic medium. Nowadays, most companies in the industry use devices that use EEG electrodes that are placed on the scalp, where the patterns of electrical activity generated when millions of brain cells are actually acting are detected. Different types of brain waves have been associated with different mental states, such as relaxation or other actions. One device to help monitor and assess stress in order to prevent unpleasant situations is to consider the Emotiv Epoc wearable helmet, which is able to record the individual's eeg signal and assess their stress level. The learner classification app was applied to five different features extracted from the whole stew dataset, containing a collection of EEG signals, acquired through wearable sensors by the subjects in two experiments, namely "NO TASK" and "SIMKAP-based multitasking activity". From the analysis of the results obtained, it can be seen that for the classification of workload levels the ensemble boosted trees model achieves 92.0% and 93.1% classification accuracy for the training and test results respectively. Consequently, it can be stated that this type of model provides good results, the remaining percentage is a good starting point for future studies, where more attention will be paid to finding more accurate methods to analyze even very noisy signals.