Purpose/Significance To promote the clinical practice of exercise training intervention in the management of heart failure (HF). Method/Process The relevant literatures, clinical guidelines, and expert consensus are systematically analyzed and integrated, and a knowledge graph (KG) of HF and exercise training is constructed. Based on the KG, a graph embedding recommendation model and a personalized exercise prescription recommendation system for the management of HF are constructed by using the fast random projection algorithm and the K-nearest neighbors algorithm.Result/Conclusion In total,the KG of HF and exercise training includes 2 703 instances and 25 161 relations. Based on the KG, the graph-embedded personalized recommendation system provides safe, effective, and diverse exercise prescription recommendations.