Collaborative filtering recommendation method based on knowledge graph embedding

Abstract

The invention discloses a collaborative filtering recommendation method based on knowledge graph embedding, and the method comprises the steps: 1, constructing a sub-knowledge graph based on user-project historical interaction and a Satori knowledge graph in a structured knowledge feature extraction stage, searching potential vector representation of structured knowledge from the sub-knowledge graph through an improved knowledge graph embedding method, and carrying out the feature extraction of the structured knowledge; obtaining a structured embedded vector so as to obtain a project representation with knowledge perception; step 2, in a joint learning stage, constructing a final project potential vector representation based on a structured embedded vector and an unstructured project feature vector, and embedding information of the structured embedded vector and the unstructured project feature vector into a unified vector space by adopting a collaborative filtering recommendation method; and step 3, in a recommendation list generation stage, using an inner product of the user vector representation and the final project potential vector representation as a preference probability value of the user, and generating a personalized recommendation list by taking the preference probability value of the user as a target.

Description

technical field [0001] The invention belongs to the field of computer applications, in particular to a collaborative filtering recommendation method based on knowledge graph embedding. Background technique [0002] In recent years, with the rapid development of network applications such as shopping, news, reading, music, video platforms and social media sites, users have to choose their favorite content from overwhelming information. In this case, recommender systems, as a tool to alleviate information overload, need to help users discover personalized interests from the ever-increasing mass of data. [0003] In order to achieve personalized recommendation, traditional collaborative filtering methods utilize various online unstructured data of users, such as clicks, ratings, etc., and predict users' preferences for items based on these historical information. In general, a learnable collaborative filtering model consists of two parts, embedding and interactive data modeling...

AI Technical Summary

Problems solved by technology

But such a knowledge base usually has a small number of project-related attributes and relationships, and most of the useful project information may be lost

Images