Epileptic Seizures Detection from EEG Recordings Based on a Hybrid system of Gaussian Mixture Model and Random Forest Classifier

Epilepsy is the most common neurological disease defined as a central nervous system disorder that characterized by recurrent seizures. While electroencephalography (EEG) an essential tool for monitoring epilepsy patients' brain activity and diagnosing epilepsy, Visual detection of EEG signal to identify epileptic seizures time-consuming approach might result in human error. Therefore, early precise diagnosis critical reducing risk future This paper aims increase seizure accuracy balanced dataset while execution time. To address this, we proposed hybrid supervised unsupervised machine learning algorithms construct computationally efficient scalable model from two-class datasets. First, Discrete Wavelet Transform (DWT) was applied decompose it into frequency sub-bands. Then these extracted features were fed Gaussian Mixture Model (GMM) partitioning two clusters: or not. Lastly, clusters' output evaluated with random forest classifier. In addition, Principal Component Analysis (PCA) used reduce further obtained after conducting DWT on determine impacts dimension reduction this performance. The experimental results show highest achieved GMM 93.62 %.