EEG-based optimization of eye state classification using modified-BER metaheuristic algorithm.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Accurate eye state classification from EEG signals is crucial for various applications.
• Existing algorithms face challenges in precision and feature selection for binary classification tasks.

Purpose of the Study:

• To introduce and evaluate the Modified Al-Biruni Earth Radius (MBER) algorithm for improved eye state classification.
• To compare MBER's performance against established optimization algorithms like BER, PSO, WAO, GWO, and GA.

Main Methods:

• EEG data preprocessing: scaling, normalization, and null value elimination.
• Implementation of the MBER algorithm for binary feature selection.
• Evaluation using multiple machine learning classifiers (KNN, DT, RF, etc.) with KNN as the optimized fitness function.
• Statistical analysis using ANOVA and Wilcoxon signed-rank tests.

Main Results:

• The MBER algorithm demonstrated superior performance on unimodal benchmark functions compared to other optimizers.
• The K-Nearest Neighbors (KNN) model, optimized by MBER, achieved high performance metrics: Precision (0.959), NPV (0.965), F-Score (0.963), accuracy (0.961), Sensitivity (0.971), and Specificity (0.950).
• The overall eye state classification accuracy reached 96.12%.

Conclusions:

• The Modified Al-Biruni Earth Radius (MBER) algorithm significantly enhances the accuracy of eye state classification.
• MBER-optimized KNN provides a robust and effective method for analyzing EEG data for eye state detection.
• The proposed algorithm shows promise for real-world applications requiring precise eye state monitoring.