The detection and classification of emotions from neurophysiological data, particularly EEG signals, has gained significant momentum in recent years, with applications extending to fields such as mental health and brain-computer interfaces (BCI). EEG provides a non-invasive method to monitor brain activity through electrodes placed on the scalp, allowing researchers to investigate how emotions manifest at the neural level. This study focuses on the classification of emotions—positive, negative, and neutral—using machine learning models in conjunction with a novel method called Higher Order Network Analysis (HONA). HONA leverages advanced statistical analyses of EEG signals to uncover features with neurobiological relevance, making it a powerful tool for detecting emotional states. Utilizing a publicly available EEG dataset, we demonstrate that our approach achieves superior classification performance compared to existing methods. This work emphasizes the critical role of feature selection in improving emotion detection accuracy and advocates for the ethical considerations necessary in deploying such technologies in real-world applications. Our findings suggest that HONA offers a promising avenue for future advancements in BCI and emotion detection research.