Recommendation method and device

Abstract

The invention relates to a recommendation method and device. The recommendation method comprises: obtaining to-be-recommended data, wherein the to-be-recommended data comprises multiple pieces of user data, multiple pieces of project data, a user knowledge graph constructed based on the multiple pieces of user data and a project knowledge graph constructed based on the multiple pieces of project data; inputting the to-be-recommended data into a trained recommendation model to obtain a preference score prediction value of each user data to each item data in the to-be-recommended data; and obtaining at least one item data for each user data as recommended item data based on the preference score predicted value of each user data for each item data in the to-be-recommended data. According to the recommendation method and device disclosed by the invention, the TransR model is adopted as a spatial conversion network in the recommendation model, correct knowledge expressed in a triple form is effectively expressed, and a more efficient fusion mode of collaborative information and knowledge propagation is realized.

Description

technical field [0001] The present disclosure relates to the technical field of big data, and more specifically, to a recommendation method and device. Background technique [0002] Knowledge graphs have been widely studied and applied in recommender systems as auxiliary information. In the application of the recommendation system based on knowledge graph technology, there are many models used for training. Among them, the RippleNet model propagates users' latent preferences along the connections in the knowledge graph. The KGCN model and the KGNN-LS model use graph convolutional networks to obtain item embedding representations through neighbor entities in the knowledge graph. The KGAT model introduces a collaborative knowledge graph (CKG), which combines user-item interaction graphs and knowledge graphs, and recursively propagates on the CKG through a graph convolutional neural network. This method assumes in advance that the items in the user-item interaction graph and...

AI Technical Summary

Problems solved by technology

[0003] The recommendation system CKAN (Collaborative Knowledge-aware Attentive Network for Recommender Systems) of the collaborative knowledge embedding attention network in related technologies cancels the user's independent embedding representation, and uses the embedding representation of the items that the user has interacted with to form the user's embedding representation. CKAN serves as An end-to-end model based on propagation, which has a certain recommendation accuracy, but ignores the processing of the complex semantic relationship space between entities in the knowledge graph

Images