Graph convolutional network recommendation method and device based on interlayer combination mechanism

Abstract

The invention discloses a graph convolutional network recommendation method and device based on an interlayer combination mechanism, and the method comprises the steps: firstly constructing a user-project historical interaction record set, constructing a knowledge graph of a commodity, constructing a project-commodity entity alignment set, and constructing a user-project historical interaction record set; coding and integrating the interactive behaviors of the user and the project and the knowledge graph of the commodity into a unified relational graph data structure, and generating a collaborative knowledge graph of the user and the commodity; carrying out graph convolution operation on feature information of each node in the collaborative knowledge graph through a graph convolution network, and extracting user and commodity features; and finally, calculating inner products of the user and the feature vectors of the commodities in sequence, carrying out sorting recall according to the sizes of the inner products, and finally recommending the commodities to the user. According to the method, the problem of excessive smoothness in a traditional graph-based convolutional network recommendation method can be effectively relieved, and meanwhile, the method has relatively high generalization ability, so that the accuracy of a recommendation result is improved.

Description

technical field [0001] The invention belongs to the technical field of commodity recommendation, and in particular relates to a graph convolution network recommendation method and device based on an inter-layer combination mechanism. Background technique [0002] Many data in real life do not have a regular spatial structure, such as graph data in non-Euclidean spaces such as transaction flow, social network, and molecular structure. The adjacent structures of each node in these graph data structures are different. In the above graph data, the feature information and structural information of nodes need to be considered at the same time. If only manual rule extraction is used, a lot of hidden and complex pattern information will be lost. Graph Convolutional Network (GCN), as a method for deep learning of graph data, converts the graph data structure into a canonical standard representation by formulating corresponding strategies on the nodes and edges in the graph. Compare...

AI Technical Summary

Problems solved by technology

[0003] Although the graph convolutional network has a strong feature extraction ability on graph-structured data, as the number of convolutional layers and the number of iterations increase, the node features in the same connected component of non-Euclidean data will tend to converge to the same value, and then Losing the characteristic information of the node itself, this phenomenon is called over-smoothing problem (over-smoothing)

Images