Glioma prognosis remains a significant challenge due to tumor heterogeneity and complex patient-specific factors. To address this, we propose an advanced diagnostic framework integrating machine learning (ML), DL, and XAI to ensure both accuracy and interpretability. The framework incorporates multiple feature selection strategies, including Important Feature Selection (IFS), Mutual Information (MI), Principal Component Analysis (PCA), and Pearson Coefficient, to extract critical tumor attributes. Patient-centric data such as medical history and genetic information are combined to enhance predictive modeling. Various machine learning algorithms and ensemble approaches are employed, with the Voting Classifier showing superior performance using Important Feature and Mutual Information techniques. Furthermore, CNN models demonstrated strong predictive power, achieving 91% accuracy with PCA-based features and 90% with Pearson Coefficient features. By coupling predictive precision with interpretative insights, the proposed approach empowers clinicians with transparent, data-driven support for improved glioma grading, diagnosis, and personalized treatment planning.