A Graph Neural Network Link Prediction and Recommendation Method Based on Social Relations

Abstract

The invention discloses a graph neural network link prediction and recommendation method based on social relationships, which establishes a user-item bipartite graph, and constructs an item interaction topological graph containing user social relationships according to the social relationships between users. The comment text and purchase relationship are used as the source of node information, and Bert is used for feature extraction of text data and network structure to obtain the initial feature vector of each node. By using the residual connection on the GNN, the node information in the graph retains the structural information of the original graph as much as possible, and the updated node information is obtained. Finally, the user's preference for the item is obtained through the link prediction algorithm, and according to the obtained prediction score, Top‑n recommendation is used to generate a list of recommended items. The present invention provides personalized descriptions of node features through comments, and can make full use of the user-item structural information of the topological graph, thereby making more effective recommendations.

Description

Technical field [0001] The present invention relates to a recommendation system, text classification and GNN research techniques, in particular to a graph neural network link prediction recommendation method based on social relationships. Background [0002] With the continuous development of Internet technology, the amount of information on the Internet has increased sharply, and it is often difficult for users to choose products that are truly suitable for themselves, so the demand for personalized recommendation services is becoming more and more extensive. How to make full use of the user's existing comments and behavior data, dig out user preferences and product characteristics, and recommend suitable products for users is very important to improve user experience and increase product sales. Therefore, the recommendation system has become a hot area in the current development of the Internet. [0003] Traditional collaborative filtering recommendation technology, generally u...

AI Technical Summary

Problems solved by technology

[0005] (1) Only relying on a single rating data, the review information is not fully utilized;

[0006] (2) User social information is rarely used

Images