Human Emotion and Mental Health Monitoring via EEG: A Fuzzy Rule-Based Machine Learning Approach

In this research, we introduced a mixed approach using machine learning models to classify emotional and psychological disorders using EEG brainwave data. The challenge of monitoring these complex and subjective states is significant. To tackle this challenge, we utilize fuzzy logic to compute a health score by combining outputs from different monitoring models. A recent study built upon our approach by introducing a multi-layer stacking classifier incorporating data from various classifiers. We applied linear regression-based correlation scores to select 997 features out of 2548, resulting in an impressive 98.75% accuracy in emotion recognition. This technique particularly emphasizes the identification of negative emotions and utilizes K-means clustering (with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k=3$</tex>) for labeling, along with a genetic algorithm from the DEAP framework for feature selection. Further optimizations included dataset balancing and hyperparameter tuning, which integrated a gradient-boosting tree classifier achieving an unprecedented 97.21% accuracy in recognizing psychological disorders. These trends indicate major developments in EEGbased emotion analysis and psychological disorder diagnosis, with potential implications for treatment. Our Fuzzy logic is used in the suggested approach to compute a health score by amalgamating outputs from emotional and psychological disorder monitoring models. The system's effectiveness is rigorously assessed using various metrics such as accuracy, F1 score, precision, recall, ROC curve, and misclassification samples. This multimodal health monitoring system holds the potential for informed decision-making and treatment planning, making significant contributions to health informatics.