Brain tumor is a very severe neurological disease, which de- mands fast and accurate therapy because of the com plicated nature of brain structures. The development of brain cancer stems from an uncontrollable growth of cells within the brain tissue. A proposed method applies a deep learning algorithm to identify brain tumors in magnetic resonance imaging (MRI). For accurate detection of brain tumors, a pre- trained model VGG16 was used through transfer learning to improve the accuracy. The dataset consists of 7023 MRI images (512 × 512 dimensions). To enhance the model performance, preprocessing steps such as normalization and data augmentation. After all these steps, the VGG16 model alone achieves an accuracy of 98.06%. To increase the performance level, the ensemble method is used i.e. averaging and stacking and there- fore, stacking achieves an accuracy of 100% and a 0.9991 AUC-ROC score. The output proves that integrating transfer learning with ensemble models increases the classification performance and accuracy of the model for the accurate brain tumor diagnosis.