Epilepsy is a neurological disorder that affects approximately 1% of the world's populations. Epilepsy prediction has been of great interest as it can identify and warn of an upcoming seizure, and to reduce the burden of the unpredictability of seizures. In this paper, we proposed an improved seizure prediction model, SimuB_ResNet, based on the pretrained ResNet, using intracranial EEG signals. We designed a simulation block to convert EEG data into image like data before the pretrained network. Since the data is extremely imbalanced, we used an improved focal loss (FL) instead of the cross entropy loss and find the relationship between the FL's parameters and the class weight. Compared with a state-of-art CNN model, our proposed model achieved better AUC of 0.89. Moreover, our results demonstrated that EEG signals can be migrated to the image network which is pretrained on large data set through a simulation block, and better results can be achieved.