Alzheimer’s disease is the most common form of dementia, affecting a significant proportion of the elderly. Early diagnosis and grouping of AD is important because in this domain, slowing down the progression of disease condition improves patient outcomes. This project focused on building and evaluating a variety of machine learning and deep learning models for the classification of Alzheimer’s from the Alzheimer MRI Preprocessed Dataset from Kaggle. The dataset includes MRI scans that capture structural brain alterations caused by the disease. Diverse classes are encountered in this dataset: non-demented, very mild demented, moderate demented, mild demented. The project will study various AI-based models for the classification of Alzheimer’s, including CNN architecture models themselves (custom-built) and pre-trained models such as ResNet, EfficientNetB0; classical machine learning models including Logistic Regression, SVM, and Random Forest; hybrid models that combine CNN with algorithms such as XGBoost, GNB, and SVM. The relevant metrics to use in order to compare each model include accuracy, precision, recall, and AUC. In general, the highest accuracies were found in conventional CNN and PCA-SVM models. Hence, through comparative analysis of the results of these models, the project will be able to determine which of them is the best model in diagnosing the different stages of Alzheimer’s disease and may prove useful in making early prognosis and management of the disease possible. Thus, this work shows how a newly trained CNN model can outperform the default model, with the accuracy of 99%, and also presents several models consisting of CNN features integrated with the more conventional classifiers. Unlike prior works, it gives a new idea and asset to enhance the detection of Alzheimer’s disease. Such improvements contribute markedly to early detection and prediction of the disease progressed aggressively.