A Novel WOA-DTW Clustering Framework for Enhanced EEG Classification Across Multi-Dataset Brain Signal Analysis

Background Accurate electroencephalogram (EEG) classification is hindered by the nonlinear, nonstationary nature of brain activity and high inter-subject variability. Conventional clustering often fails to capture the temporal and complex EEG patterns, limiting cross-dataset performance. Method This study introduces a hybrid framework combining the Whale Optimization Algorithm (WOA) with Dynamic Time Warping (DTW) to optimize cluster centroids using temporal similarity, enhancing EEG dynamic representation. The method was validated on four real-world EEG datasets covering binary and multi-class tasks. Preprocessing and dimensionality reduction preceded clustering, followed by evaluation using five different classifiers. Results WOA-DTW consistently outperformed baseline and traditional clustering methods, achieving accuracy gains up to 13% and F1-score improvements of 15%. The highest accuracy, 94.98%, was obtained on dataset 2 (EEG SSVEP) using CNN. Paired t-tests confirmed statistical significance (p < 0.05). Compared with K-means, DBSCAN, FCM, SOM, and standalone DTW, the proposed method better captured nonlinear and temporal EEG characteristics, improving classifier discriminability. Conclusions The WOA-DTW framework effectively models EEG temporal dynamics, significantly enhancing classification performance and showing strong potential for brain–computer interfaces and neurological diagnostic applications.