Personalized recommendation method based on attributes and scores

The invention discloses a personalized recommendation method based on attributes and scores. The personalized recommendation method comprises the following steps: S1) calculating similarities among users; S2) calculating the similarities among all items; S3) establishing a user similarity matrix; S4) establishing an item similarity matrix; S5) establishing a scoring matrix for a most similar item by a most similar user; S6) taking the scoring matrix for the most similar item by the most similar user as the input of a BP (Back Propagation) neutral network for carrying out training; S7) judging whether the BP neutral network is subjected to convergence or not, and entering an S8) if the BP neutral network is subjected to convergence, otherwise, carrying out BP neutral network parameter and function revision on the BP neutral network, and returning back to the S6); and S8) utilizing the BP neutral network to carry out operation on a test set to generate a final recommendation result and recommend the final recommendation result to a target user. The personalized recommendation method can carry out accurate relevant recommendation to a new user who joins in the system, and meets the recommendation requirements of the user.