Early detection of Alzheimer's disease (AD) is essential. Compared with manual judgment, using machine learning to recognize AD has the advantages of high efficiency and low cost. After reviewing the research progress in using machine learning technology and neuropsychological data to identify AD, we point out that, firstly, the dataset of AD should be expanded by discussing around a specific topic, secondly, more reasonable features should be automatically selected by machine learning, at last, DNN model for predicting AD should be further study.