Among neurological disorders, epilepsy is the most sporadic and recurrent. In cases where a patient is in a rural area, early diagnosis of epileptic seizures is extremely important for ensuring the prompt administration of appropriate medication. To identify seizures in an Electroencephalogram (EEG) signal, the authors of this research employ a deep learning system. The dataset used in this investigation, which included pre-ictal, inter-ictal (epileptic seizure-free), and ictal EEG signals, is openly available (epileptic with seizure). It has been suggested that the Bayesian optimization technique be utilized for the Double LSTM model to complete the EEG signal categorization objective. The proposed double LSTM classifier achieves up to 95% accuracy in categorizing. These are three categories of signals, with 96% of balanced specificity and sensitivity. The feature extraction procedure uses the VGG-16 model. Double LSTM architecture with recommended Bayesian optimization improves classification accuracy by 6.52% over the SVM model and 3.15% over LSTM.