Performance Evaluation of Integrated Deep Learning and Ensemble Methods for Epileptic Seizure Detection

Automated seizure detection from EEG data is essential for improving the quality of life for individuals with epilepsy. This study evaluates multiple machine learning and deep learning models to identify the most accurate and efficient approach for seizure detection across two distinct EEG datasets. The models analyzed include a novel Voting Classifier ensemble (combining SVM, Random Forest, and XGBoost), CNN, DWT-based DNN, MDBCN, and standalone SVM, Random Forest, and XGBoost classifiers. Among these, the Voting Classifier consistently demonstrated superior performance, achieving 100% accuracy, precision, recall, and F1-scores with competitive computation times of 9.63 seconds and 15 seconds on the two datasets, respectively. Other models showed strong performance, but with notable limitations, such as high computational demands or reduced recall in certain cases. These findings highlight the Voting Classifier's balance of accuracy and efficiency, establishing it as a reliable solution for automated seizure detection. Future research will focus on optimizing computationally intensive models, exploring hybrid approaches, validating on diverse datasets, and integrating these techniques into real-time systems to enhance patient care.