Gliomas are the most frequent malignant brain tumors and accurate quantification is vital for the planning of treatment as well as for determining prognosis. However, currently existing machine learning (ML) models suffer from high dimensionality of the data as well as imbalanced distribution of classes making these models particularly difficult to deploy in a general way. This study enhances ML algorithms for glioma grading and mutation classification by optimizing feature selection, employing hybrid models, and leveraging deep learning techniques. Experimental results show improved accuracy and computational efficiency, demonstrating the potential for more reliable, automated glioma diagnostics in clinical settings. The dataset used represents glioma grade, indicating severity or progression of the tumour. It comprises of 839 records and 24 columns representing patient information related to glioma grading and genetic mutations. These predictor attributes include demographic factors such as gender, age and race. The records were labelled based on grade which represents glioma grading i.e. low-grade gliomas and high-grade gliomas. Various ML methods like trees, logistic regression, SVM, naïve bayes, KNN and ensemble were utilized. The performance metrics of all the models were compared. The findings indicate that the Ensemble and SVM show the best validation accuracy of 87.5% and 87% in classifying glioma grades.