Ensemble Fusion Models Using Various Strategies and Machine Learning for EEG Classification.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Electroencephalography (EEG) is a low-cost method for assessing brain electrical activity.
• Accurate classification of EEG signals is crucial for diagnosing neurological disorders like epilepsy.
• Machine learning and deep learning offer automated EEG signal classification.

Purpose of the Study:

• To propose and evaluate five novel ensemble models for automated EEG signal classification.
• To identify the most effective model for diagnosing neurological disorders, with a focus on epilepsy.

Main Methods:

• Development of five distinct ensemble models incorporating techniques like ESCD feature selection, Infinite Independent Component Analysis (I-ICA), Genetic Algorithm (GA), Hilbert Huang Transform (HHT), and Factor Analysis.
• Utilizing Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) classifiers within ensemble frameworks.
• Comparative analysis of model performance for EEG signal classification.

Main Results:

• The proposed ensemble hybrid model, combining equidistant assessment, ranking determination, ESCD feature selection, and SVM classification, achieved the highest accuracy.
• This leading model demonstrated a classification accuracy of 89.98% for EEG signals.

Conclusions:

• Ensemble modeling approaches significantly enhance EEG signal classification accuracy.
• The developed ESCD-based feature selection technique combined with SVM offers a promising method for automated epilepsy detection via EEG.