<p>Accurate classification of brain tumors is crucial for effective diagnosis and treatment. One of the challenging problems is the multi-class classification of brain tumors using MR images. However, brain tumor classification often suffers from limitations related to interpretability, feature extraction, and classification accuracy. To address these issues, the proposed model employs three parallel convolutional branches, each designed to extract complementary multi-scale features from MRI images. Specifically, 1 × 1, 3 × 3, and 5 × 5 convolutional kernels are used in separate branches to capture fine-grained intensity variations, intermediate texture patterns, and global structural context of brain tumors, respectively. The multi-scale features extracted from all branches are fused via feature-level concatenation, followed by fully connected layers, and the resulting 256-dimensional feature representation is classified using a Random Forest classifier to enhance robustness and interpretability. Furthermore, to overcome the preprocessing data limitation, such as brightness and contrast variations in brain MRI images, we employed an efficient preprocessing strategy that normalizes the images, reduces noise, and enhances image quality. The DCNN component of the model performs feature extraction through convolutional layers, learning both low-level and high-level features, while parallel pathways are introduced to capture diverse features from different aspects of the input data. The extracted features are then fed into the Machine Learning Model (MLM), which refines the classification using interpretable and context-aware decision-making mechanisms. The proposed model efficiently classifies brain tumors into three categories, i.e., glioma, meningioma, and pituitary tumor. The experimental results demonstrated that the proposed model outperforms existing methods in terms of performance metrics with an accuracy of 98.6%.</p>