Purpose/Significance To construct the identification model of potentially highly cited papers, to excavate the research frontier, so as to provide new ideas for the research of frontier identification. Method/Process The paper determines the feature system from three dimensions, builds the identification model of potentially highly cited papers based on six machine learning algorithms, selects the best model to identify potentially highly cited papers, and uses clustering and other methods to mine the research frontier of the selected literature datasets.Result/Conclusion Taking the field of artificial intelligence medicine for Alzheimer’s disease as an example, the XGBoost model has the best performance evaluation results. Four types of research frontier topics have been effectively identified:intelligent auxiliary tools, early disease prediction and classification, biomarkers, disease risk assessment and care.