Pancreatic cancer has high mortality rates because it is often not diagnosed early and is typically discovered at a later stage when there are few available treatments. This underscores the urgent need for automated systems to detect cancer early, which can significantly improve diagnosis and treatment outcomes. While various algorithms have been used in the medical field to help with early detection, there is still significant potential for developing more advanced computational systems. The main goal of this study is to improve pancreatic cancer detection utilizing hybrid learning methods. Specifically, this research focuses on creating a system that examines medical imaging data, mainly CT scans, to find essential features and cancerous growths in the pancreas using models based on deep learning. Convolutional neural networks (CNN), transfer learning model (VGG-16), and recurrent neural networks base LSTM were applied in the study. Initially, every model was applied individually and later, with the combination of CC and LSTM along with Vgg-16, we proposed a hybrid model. A novel optimization approach was used to boost the hybrid model performance. The study aims to assess the effectiveness of these models, particularly the innovative Hybrid approach, in identifying pancreatic cancer. Our results show that the presented hybrid model outperforms other methods, achieving a testing accuracy of 98%. The precision, recall, and f-score accordingly 98.3%, 98.0%, and 98.1%. This demonstrates that the hybrid model shows promise in detecting pancreatic cancer. The proposed model outperformed the state-of-art models.