Collaborative relation graph-based recommendation method and related device

Abstract

The invention discloses a recommendation method based on a collaborative relation graph and a related device. The recommendation method comprises the steps that the collaborative relation graph is constructed, interaction nodes with entity nodes are obtained based on the collaborative relation graph, and implicit relation embedding representation of the entity nodes is determined based on the interaction nodes. Obtaining triples taking the entity nodes as head entities on the basis of the collaborative relation graph, and determining explicit relation embedding representation of the entity nodes on the basis of all the obtained triples; obtaining high-order domain information of the entity node based on the implicit relation embedding representation and the explicit relation embedding representation; and training a recommendation model based on the high-order domain information, and determining an interaction probability between the to-be-recommended user and each candidate item corresponding to the to-be-recommended user through the recommendation model. According to the method, the user-article bipartite graph and the knowledge graph are integrated, and meanwhile, the implicit relationship between the user and the article and the attribute relationship between the entities are utilized to carry out high-order information propagation, so that the high-order neighborhood information of the nodes is effectively explored, and the recommendation accuracy is improved.

Description

technical field [0001] The present application relates to the field of information technology, in particular to a recommendation method based on a collaborative relationship graph and a related device. Background technique [0002] With the rapid development of the Internet era, it has become difficult to quickly and accurately obtain the required information from massive resources. This problem is called information overload. As a means to effectively alleviate information overload, recommender systems have important practical research value. Traditional recommendation algorithms such as collaborative filtering (Collaborative Filtering) methods can build a "user-item" interaction matrix through the user's historical behavior, thereby recommending items that similar users like to target users, or recommending items that target users like. of similar items. However, on the one hand, the collaborative filtering method has the problem of data sparsity, that is, it does not pe...

AI Technical Summary

Problems solved by technology

However, the current recommendation method of knowledge graph only utilizes the direct relationship between entities, and ignores the potential relationship between "user-item", which makes it have the problem of poor ability to discover high-order neighborhood information in the graph model. resulting in poor recommendation results

Images