Segmentation of brain tumor from the medical imaging data plays a significant part in the diagnosis, treatment planning, and monitoring of neuro-oncological diseases. The current study focuses on exploring the effectiveness of machine learning models, such as modified Support Vector Machine (SVM), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and VGG 16, in improving the accuracy of brain tumor segmentation. The diverse data set includes 4320 brain tumor images from CT and MRI datasets that have been collected and preprocessed to standardize picture quality and eliminate noise. Features were obtained using the Berkeley Wavelet Transform to capture a functional picture of the spatial expressions and frequencies of tumor-tissue in the diverse data set. Four models were produced from the acquired images, and the efficacy of the characteristics in representing the brain tissue as tumor and non-tumor tissues was assessed. In all the produced models, 70% of the dataset was used to train the features, and the rest 30% was utilized to test the machine learning facilitated network. It was discovered that the Modified SVM model had the highest accuracy at 98.65%, while VGG 16 represented the second highest accuracy 95.45%. CNN had 93.4% accuracy while RNN was the fourth at 91.2%. The precision, recall, and F1 score were analyzed in all the models, and the study produced confusion matrices to observe how the functions performed in classifying brain tumor and non-brain tumor tissues.