EmotionRF: Enhancing Emotion Classification from EEG Signals Using Machine Learning with Explainable AI

This study investigates the application of machine learning techniques for classifying human emotions using electroencephalography (EEG) data. Focusing on the categorization of positive, negative, and neutral emotional states, we utilize the ‘Feeling Emotions’ EEG brainwave dataset from Kaggle to evaluate the efficacy of various classification models.Seven machine learning algorithms are employed in this analysis: Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Gaussian Naive Bayes (GNB), ExtraTrees (ET), and Decision Tree (DT). Our proposed model, ‘EmotionRF’, based on RF, achieves 99% accuracy, significantly outperforming other algorithms. To enhance the interpretability of the high-performing Random Forest model, we apply Explainable AI techniques, specifically LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanations). These methods provide valuable insights into the model’s decision-making processes and feature importance, thereby increasing the transparency of its predictions and confirming its robustness.Our findings highlight the potential of EEG data in developing reliable tools for psychological assessment and adaptive interfaces. This research contributes to the advancement of machine learning applications in neuroscientific data analysis and underscores the importance of interpretable AI in clinical and research settings.