Accurate Epileptic Seizure Detection from EEG Using Feature Fusion and MI-Enhanced XGBoost Classifier

The accuracy of Electroencephalogram (EEG) signal-based epileptic seizure identification is often compromised by poor feature selection and duplicate data. This research proposes a method that combines early feature fusion from many domains with Mutual Information (MI)-based feature selection to overcome these issues. Principal Component Analysis (PCA), Hilbert–Huang Transform (HHT), Reconstruction Independent Component Analysis (RICA), and Empirical Mode Decomposition (EMD) are used to extract features that capture time, frequency, and nonlinear information. The Extreme Gradient Boosting (XGBoost) algorithm is used to categorize the most relevant qualities once Mutual Information has been utilized to choose them. The suggested approach performs exceptionally well on all significant measures when using the Bonn EEG dataset. Its efficient design ensures both enhanced detection capability and suitability for real-time clinical use.