Epilepsy is a malfunction of nervous system that it changes the quality of human’s life, and it is caused by many factors. Therefore, the diagnosis of this disease needs the acquisition of the electrical activity of brain which called electroencephalography (EEG) technique. Thus, EEG signals (EEGs) recorded are widely used for epileptic seizure detection and prediction. However, to investigate these process, convolutional neural network (CNN) and deep recurrent network (DRN) are developed to automate the classification of EEGs. In the presented article, a proposed model combined both one-dimensional CNN (1D- CNN) and Gate Recurrent Unit (GRU) is established for EEGs analysis. Furthermore, the performance of this framework is evaluated by using the publicly available CHB-MIT dataset, and the classification between these conditions is examined. Moreover, the obtained results show higher accuracies of the proposed 1D-CNN-GRU and outperforms significantly previous works. Finally, this architecture is accurate and proves their efficacity in the task of epileptic seizure detection and prediction.