Novel Multi Block-Segmented Feature Selection Approach Using the Szymkiewicz–Simpson Coefficient with PSO for High-Dimensional EEG Classification with Machine Learning in Edge Computing Systems

Electroencephalogram (EEG) signal classification plays a vital role in medical diagnosis, cognitive analysis, and human-computer interaction.However, the high dimensionality and redundancy of EEG features often degrade the accuracy and efficiency of machine learning models, especially in resource-constrained edge environments.To address these challenges, this study introduces a novel multi block-segmented feature selection approach that integrates the Szymkiewicz-Simpson (SS) coefficient with Particle Swarm Optimization (PSO) to identify the most relevant and discriminative EEG attributes.In the proposed SS-PSO framework, the first stage divides the high-dimensional EEG feature space into multiple blocks (ranging from 50 to 250 samples per block-feature) to compute SS similarity scores and eliminate low-relevance features, thereby reducing the search space.The second stage applies PSO to refine the remaining high-harmony features and select an optimal, compact subset.The approach was validated using two highdimensional EEG datasets: EEG Brainwave and Bonn EEG.The number of features was reduced from 2549 to 892 (~65% reduction) in the EEG Brainwave dataset and from 4097 to 1639 (~60% reduction) in the Bonn EEG dataset.Using the optimized feature subsets, the K-Nearest Neighbors (KNN) classifier achieved 98.91% accuracy on the EEG Brainwave dataset and 99.33% accuracy on the Bonn EEG dataset.These results demonstrate that the proposed SS-PSO feature selection method significantly enhances classification accuracy while reducing computational complexity, making it suitable for real-time and low-power EEG analysis in modern edge computing environments.