Xavier-PSO-ELM-based EEG signal classification method for predicting epileptic seizures

Epilepsy represents one of the most common neurological diseases that affects a substantial number of individuals worldwide, which is characterized by recurrent, unprovoked seizures detectable via electroencephalogram (EEG). To address this issue, we propose an Extreme Learning Machine (ELM) model for seizure prediction. Firstly, we optimized the ELM with a single hidden-layer feed-forward network using Particle Swarm Optimization (PSO) as a meta-heuristic. Secondly, we employed the Xavier method to initialize random variables and improve model performance. Our optimized classification model was evaluated using a dataset from the University of Bonn. Results from the experiments demonstrate our model's excellent classification performance with 98.13% sensitivity and 91.04% specificity.