The invention discloses a label information-considered Bayesian personalized ranking recommendation method. The method is characterized by comprising the following steps of: 1, defining an interactionrelationship between users and products; 2, defining a label matching degree between users and interaction item products; 3, defining a division standard for preference feedback sets of the users; 4,constructing a matrix decomposition model, for the products, of the users; 5, solving the model by utilizing a Bayesian personalized recommendation method; and 6, obtaining descending ranking, for the interaction item products, of a certain user and recommending the first top products to the user. The method has relatively good recommendation performance under the conditions of data sparse and cold start user recommendation.