Analysis of EEG signals for the detection of epileptic seizures using feature extraction

Abstract

The electroencephalogram, which tracks electrical signals in the central nervous system, has been extensively used to diagnose epilepsy, which represents a particular sort of brain abnormality. However, developing seizure classification techniques with significantly better precision and reduced complexity remains challenging. The Epileptic Seizure Recognition dataset, which is publicly accessible in the Kaagle and in the machine learning repository, was used to identify seizures. To identify the seizure, we compared six classification methods to determine which one had the highest success rate. The dataset is subsequently divided, trained, and tested in order to categorize it further using six machine learning algorithms: Stochastic Gradient Descent, Logistic Regression, Naïve Bayes, K-Nearest Neighbors Algorithm, Extra Tree Classifier and Decision Tree. When contrasted with alternative techniques, Extra Trees Classifier possesses the highest accuracy results. The algorithm attained a 96 percent success rate.