Emotion classification using Electroencephalogram (EEG) signals is increasingly recognized for its potential in fields such as neuro-marketing, mental health monitoring, and human–computer interaction. This study investigates the application of machine learning algorithms to classify emotions from EEG data, with a particular focus on optimizing the Support Vector Machine (SVM) algorithm. The dataset utilized includes EEG brainwave data from two subjects, recorded across three emotional states—positive, neutral, and negative—using a Muse EEG headband at TP9, AF7, AF8, and TP10 electrode placements, supplemented by six minutes of neutral resting-state data. Five machine learning algorithms were employed: SVM, Naive Bayes Classifier, Logistic Regression, Decision Tree Regressor, and Random Forest. Optimization of the SVM involved a detailed grid search to fine-tune the kernel function, kernel coefficient, and regularization coefficient, followed by a random search to ascertain the most effective parameter combination. The study demonstrated that the SVM, particularly with a Radial basis function (RBF) kernel, outperformed other algorithms, achieving an initial accuracy of 95.78% and enhancing it to 97.66% through optimization. Cross-validation confirmed the model's robust generalization across different subsets, verifying the SVM’s enhanced performance and stability. This research highlights the efficacy of optimized SVMs in EEG-based emotion recognition, providing a benchmark for future algorithmic enhancements in this domain.