This study addresses the challenge of intelligent diagnosis in Alzheimer's Disease by employing machine learning to classify MRI images depicting various disease stages. The author conducted a comparative analysis to assess the efficacy of diverse models for this purpose, yielding several key findings. The primary model utilized in the research is founded on ResNet152V2, encompassing 152 pre-trained layers, 1 GlobalAveragePooling2D layer, 3 dropout layers, 3 dense layers, and 1 output layer. The research aims to achieve precise classification of MRI images corresponding to distinct Alzheimer's Disease stages, a pivotal step for early detection and intervention. To achieve this, this research adjusted the structure of the original model and changed some of the parameters and compared their results. Based on the results obtained, the author drew several conclusions. The author found that setting a larger value for the "patience" parameter in the Early Stopping callback helps avoid stopping the training process too early. Additionally, the author observed that the original model with only 2 fully connected layers was not stable enough in terms of gradient descent, indicating the need for additional fully connected layers. In conclusion, this research paper addressed the problem of classifying MRI images of different stages of Alzheimer's Disease using machine learning techniques. The comparative study of different models provided insights into the performance and potential improvements of the classification taskt.