This paper addresses the increasingly important role of EEG brain signals in the diagnosis and treatment of degenerative neurological and mental illnesses and disorders. The analysis and classification of these signals are essential for supporting accurate diagnosis of brain diseases and for enhancing understanding of cognitive processes. Automated classification methods for brain and EEG are vital to ensure proper assessment and treatment of neurological disorders, as manual classification can be time-consuming, error-prone, and expensive. The aim of this thesis is to focus on the classification of brain EEG signals in the two most important areas of epilepsy. A model composed of three stages is proposed: feature extraction, selection of the strongest features, and final classification. Waveletbased feature extraction is employed, using three statistical functions to choose the most informative features. Six supervised machine learning techniques, including cosine similarity, are then applied to classify the EEG signals. The results show that the neural network achieved the highest accuracy of 100%, followed by random forest and decision tree, while the k-nearest neighbor algorithm produced the lowest accuracy. This study provides insights into the effectiveness of different machine learning techniques in classifying EEG signals, which can contribute to the development of more accurate and efficient diagnostic and treatment methods for neurological disorders.
EEG signals, epileptic seizure, cosine similarity, machine learning, the probability density function (PDF)
Cite this paper
Ahmad Al-Qarem. (2023) Automated EEG Classification Using Machine Learning Approaches. International Journal of Biology and Biomedicine, 8, 20-26