Brain tumor classification is a challenging task due to the complexity of tumor characteristics, in tumor shapes, and limitations in image acquisition techniques in CNN. In this study, the Transfer Learning (TL) model is used to improve network classification accuracy and performance when applied to images of brain tumors. Also, to enhance the optimal result in classification, the TL hyperparameters are fine-tuned by the AQ- based Metaheuristics strategy. The results showed that, in terms of accuracy, sensitivity, and specificity, the proposed method performed better than other popular existing methods with a high exactness of 97.14 percent. The proposed method is evaluated and tested in brain tumor MRI dataset image.