Currently, a growing number of medical diagnostic applications feature computerized flaw detection in medical imaging. It is very crucial to find tumors in Magnetic Resonance Imaging (MRI), as this technique offers valuable information about aberrant tissues and facilitates treatment planning. Human inspection has been typically the main method for MRI brain image fault detection, but it is problematic when dealing with extensive volumes of data. So, to save time, Self-acting tumor identification techniques have been created because tumors are complex and diverse, finding them in MRI brain scans is a difficult undertaking. To find tumors in brain MRIs, we use machine learning techniques in this study. The three primary components of the suggested method are preprocessing of brain MRI images, feature extraction using Histogram of Oriented Gradients (HOG), feature selection using Cuckoo Search Optimization (CSO) algorithm, and classification using machine learning methods (KNN, XGBoost, and Random Forest).