In order to categorize brain tumors, this study uses a deep convolutional neural network (DCNN) based on Henry gas bird swarm optimization (HGBSO). The Henry gas solubility optimization (HGSO) and bird swarm algorithm techniques were merged to form the HGBSO algorithm, which was used to train the DCNN classifier. Following the first preprocessing of the images using the Gaussian filter, the Region of Interest extraction approach is utilized to reduce noise from the input MR images. After that, the regions of brain tumors are divided using a deep fuzzy clustering technique. Meanwhile, key characteristics are taken from the segmented image in order to perform an efficient classification procedure. Moreover, for improving the classification accuracy rate, data augmentation is performed. Finally, the augmented data along with total features are considered as input for developed HGBSO trained DCNN classifier, where the classification of brain tumor is performed. In terms of various metrics, the developed strategy performs better than other existing methods, obtaining values of 0.9221 for accuracy, 0.9324 for sensitivity, and 0.9295 for specificity.