Classification of Epileptic EEG Signals Using DWT-Based Feature Extraction and Machine Learning Methods

Abstract

Epileptic attacks can be caused by irregularities in the electrical activities of the brain. Electroencephalography (EEG) data demonstrating electrical activity in the brain play an important role in the diagnosis and classification of epileptic attacks and epilepsy disease. This study describes a method for detecting epileptic attacks using various machine learning methods and EEG features obtained with the Discrete Wavelet Transform (ADD). In the study, an EEG dataset consisting of five separate clusters from healthy and sick individuals was used, and the classification success between these conditions was examined separately. Support Vector Machine (SVM), Artificial Neural Networks (ANN), k-Nearest Neighbor (k-NN), Decision Trees (Tree), Random Forest, and Naive Bayes machine learning methods, which are widely used in classification, were used. In addition, comparisons were made with various windowing and overlap ratios. As a result, classification successes, as well as optimal windowing and overlap ratios were determined for various EEG clusters in the dataset.