Abstract :In recent years, the idea of using deep learning-based systems for the detection and treatment of various diseases has attracted a lot of attention. Epilepsy is a common neurological disorder worldwide. Although various methods have been developed for the diagnosis of this disease, which affects daily life in many ways and negatively, definite success has not been achieved in this regard. The idea of using deep learning methods for the detection of epilepsy is promising for the future. In this study, preictal, ictal and postictal states of intracranial EEG signals were classified by applying five different classification architectures to low frequency scalograms in a Convolutional Neural Networks (CNN) based system. These classifiers are XGBClassifier, GaussianNB, LinearSVC, Random forest classifier, SGD Classifier. The frequency of EEG signals contains important information. EEG has become an indispensable method to detect pathological conditions and anomalies and to examine the brain activities of healthy individuals. Model-adapted wavelet transform is used to improve feature selection, hypothetical testing, and classification. Epileptic IEEG data from 16 patients with a dataset sampling frequency of 512 Hz were used. Accuracy and F1-score parameters were used for experimental studies. The results obtained are 94.54% accuracy and 94.53% F1-score value for XGBClassifier. Keywords : Epilepsy, CNN, Electroencephalogram (EEG)
