Session social recommendation method based on dynamic routing graph network

Abstract

The invention discloses a session social recommendation method based on a dynamic routing graph network. According to the method, based on a given user historical interaction sequence and a social network relationship, modeling is carried out on interests and social influences of users to recommend articles which the users are most likely to be interested in in the next step. The method is mainlycomposed of five parts: in the first part, all user sequences are divided into session sequences, session vector representation is generated by using a long-short-term memory network, and the currentsession vector representation of a target user is current interest vector representation; in the second part, an attention mechanism is used to obtain the long-term interest related to the current interest of the target user; in the third part, the social influence of friends on the target user is obtained from the social network by using the dynamic routing graph network. and in the fourth part,in combination with the current interest, the long-term interest and the social influence of the user, the next clicked article of the user is predicted.

Description

technical field [0001] The invention belongs to the technical field of Internet services, and in particular relates to a conversation social recommendation method based on a dynamic routing graph network. Background technique [0002] Many online platforms, such as Yelp, Epinions, etc., allow users to share their interests and experiences and interact with other users on the platform. The behavior information of users on the platform is very rich, and generally speaking, the interests of users are diverse and changeable. Many social recommendation methods consider the user's short-term interest, long-term interest and social influence at the same time when recommending items, which can reduce the sparsity of data. The starting point of the present invention is to combine the user interaction behavior data and the user's social network in a complementary manner to improve the accuracy of the personalized recommendation method. [0003] Most social recommendation methods ado...

AI Technical Summary

Problems solved by technology

However, this method ignores an important fact that the influence of a friend on the target user is not only related to the friend and the target user, but also to other friends of the target user.

Images