Over the past decade, high-frequency oscillations (HFOs) have been studied as a promising biomarker for localizing epileptogenic areas in drug-resistant patients requiring pre-surgical intervention, while exploiting intracranial electroencephalographic iEEG. Consequently, it's important to develop accurate methods for predicting epileptic seizures. Seizure prediction involves classifying appropriate indicators, which is a difficult classification problem. Deep learning techniques, such as convolutional neural networks (CNN), have shown great promise in analyzing and classifying epilepsy-related iEEG signals. In this study, we proposed three global methods, multiclass SVM, multiple architecture CNN, and CNN-SVM, which are evaluated on a simulated iEEG dataset and then on a real iEEG signal. Our best results for the three models yield high accuracy rates, GoogLeNet-SVM achieves approximately 99.63% and 94.07% for simulated data (1) and real data (2), respectively, SVM multiclass achieves 98.14% and 88.51% for (1) and (2), respectively, and GoogLeNet achieves 98.52% and 91.85% for (1) and (2), respectively. Furthermore, the proposed model performs better than other current techniques. These results suggest that deep learning models could be a successful strategy for classifying epilepsy indicators and could potentially improve seizure prediction methods, thus improving the life quality of epileptic patients.