Modelling a Risk-based Network Model for Epileptic Seizure Prediction using Learning Approaches

The process of developing across space and time in the networks of a person with epilepsy occurs through Epileptic seizures. The generalizable technique is developed in this research to predict a particular patient seizure using the evaluation of featurerepresentation to obtain the features from the signals of multichannel EEG. The features are revealed for the signals of EEG using the available parameters. The features are input to the Risk-based Elman learning model (r - ELM) to evaluate feature representation to collectively train the data. The suggested model of r-ELM obtains 0. 096/h as the rate of false prediction, 85% as sensitivity, and 10% as the time in warning to perform the tests from the EEG dataset of CHB-Mn scalp using 10 patients. The suggested method has superiority over the existing results. Various metrics are used in the experiment which shows the epileptic stage as the essential factor affecting seizures’ performance. A subject-oriented method for seizure prediction is presented in the proposed system, which is powerful for the unbalanced data and created for any dataset of scalp EEG with no requirement of subject-oriented engineering.