Analysis of Electroencephalography Signals for Sentiment Classification: An Integrated Machine Learning and Deep Learning Approach

Sentiment detection is an expanding area aimed at comprehending human emotions from various data sources, including text, voice, and physiological signals such as Electroencephalography (EEG). EEG, a non-invasive technique monitoring brain activity, has gained prominence in this pursuit, employed in both clinical and research contexts to investi-gate brain responses to emotions. This approach becomes particularly valuable in limited communication scenarios, such as brain-computer interfaces and mental health. For this purpose, machine learning and deep learning are essential. Models such as Support Vector Machines (SVM), Deep Neural Networks (DNN), and Graph Convolutional Neural Networks (GCNN) have achieved success in sentiment classification. In this study, these models were explored and compared for emotion detection from EEG. The DNN model achieved a remarkable accuracy of 86.08%, outperforming SVM, albeit with significant processing time. The GCNN model, with its non-linear learning approach, achieved an average accuracy of 87.04% in the experiment, emerging as the most promising methodology for future research.