Detection of Epileptic Seizure from EEG Signal Data by Employing Machine Learning Algorithms with Hyperparameter Optimization

Abstract

Epileptic seizure refers to a brief occurrence of signs in the brain caused by abnormally high or synchronized neuronal activity. With the utilization of EEG signal, the epileptic seizure can be identified. However, incorporating machine learning classifiers with this EEG data can significantly contribute in detecting epileptic seizure in an automated manner. In this paper, nine machine learning algorithms have been studied and models have been constructed by utilizing UCI Epileptic Seizure dataset. The performances of the ML models are noted and detailed comparative analysis has been exhibited for both hyperparameter tuning and without hyperparameter tuning. Random search cross validation has been used for tuning the hyperparameters. Satisfactory results have been witnessed in terms of different performance metrics like accuracy, precision, recall, specificity, FI-Score, and ROC. After simulation, Support Vector Machine (SVM) performed the best in terms of accuracy with over 97.86%. However, Random Forest (RF) and Multi-Layer Perceptron (MLP) also depicted promising accuracies of 97.50% and 97.26% respectively. Therefore, with proper implementation of the ML based diagnosis system, the patients having epileptic seizures can be identified and treated at an early stage.