Online social platform user friend recommendation method based on multi-core graph convolutional network

Abstract

An online social platform user friend recommendation method based on a multi-core graph convolution network comprises the steps of constructing a social relation network according to online social network platform user relation data, constructing a multi-core graph convolution network model, wherein the model comprises two layers of structures, the first layer is a three-dimensional graph convolution kernel, the second layer comprises a plurality of three-dimensional graph convolution kernels, obtaining a node embedding matrix, carrying out inner product operation on the embedding matrix, adding an activation function, obtaining a node similarity matrix, arranging all node pairs without connected edges in a descending order, and taking the node pair with a relatively front arrangement anda corresponding user as friends recommended to each other. In the invention, high-order neighborhood characteristics are taken into consideration, a network is optimized by using multi-kernel convolution, and the accuracy and the prediction precision are improved.

Description

Technical field[0001]The invention relates to the field of social network user recommendation, in particular to a method for recommending friends on an online social platform user based on a multi-core graph convolutional network.Background technique[0002]Human beings are inseparable from social interaction. With the continuous innovation of network technology and the development of social software such as Facebook and Twitter, people's social needs have been further met. Social networks are designed to store the interrelationships between users in the network. Through social networks, it is possible to better analyze users' network images and understand users' interpersonal networks and people they may know. Through the research and analysis of social networks, precision marketing, content recommendation, and knowledge dissemination can be realized, which can profoundly affect people's information acquisition, thinking style and life. It is a blue ocean with great market prospects ...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the problems that the existing methods are difficult to effectively and comprehensively capture the multi-order relationship of users, and the accuracy of friend recommendation is not high, the present invention proposes an online social platform user friend based on multi-kernel graph convolution network with strong representation ability and good recommendation effect. recommended method

Images