Anomalous EEG Signal Time Series Classification Using Modified Metaheuristic Optimized RNN

This study delves into the significance of employing the electroencephalogram (EEG) as a noninvasive epilepsy detection method, emphasizing the potential for improvement through the integration of artificial intelligence (AI) techniques. The conventional EEG, while an invaluable tool, presents opportunities for enhancement in terms of cost-effectiveness, precision, and the specialized expertise required for its interpretation. The research at hand proposes the utilization of a particle swarm optimization (PSO) to optimize recurrent neural networks (RNNs) in the context of epilepsy prediction within a multivariate time series framework. By harnessing the power of metaheuristic optimizers, this study demonstrates their effectiveness in fine-tuning the hyperparameters of RNNs for the precise classification of seizures in EEG data. The results are promising, with the best performing models achieving a remarkable accuracy rate exceeding \(99.8\%\) when tested on previously unseen data. This research not only underscores the potential of AI-driven EEG analysis in epilepsy detection but also contributes to addressing the existing challenges associated with traditional EEG methodologies.