Recommender system based on graph convolution technique

Abstract

A recommendation system based on graph convolution technology, including: a preprocessing module, a heterogeneous graph generation module, a model training module, and a recommendation result generation module, wherein: the preprocessing module performs data cleaning and formatting on the interaction records between users and items Standardize the operation, and generate an interaction sequence for each user and output it to the heterogeneous graph generation module; the heterogeneous graph generation module constructs three heterogeneous graphs representing user preferences, dependencies between items, and similarities between users based on the user's interaction sequence data Graph and output the generated graph structure data to the model training module; the model training module trains the graph convolution model based on the graph structure data, and generates a vector representation for each user and item; preferences and generate the final recommendation results. The invention solves the problem that the number of neighbors of each node is not equal, uses the convolution operation to mine the information of the neighbors of the nodes in the heterogeneous graph, and improves the recommendation effect.

Description

technical field [0001] The present invention relates to a technology in the field of information processing, specifically a system that uses graph convolution technology to mine interaction data between users and items to implement personalized item recommendation for users. Background technique [0002] Recommended system. According to the type of data used, recommender systems can mainly be classified into content-based recommendation and collaborative filtering recommendation. The former is based on the personal information of the user and the content information of the item to model the recommendation system. Collaborative filtering technology is based on the historical records of user interaction with items to model users' preferences for items. In general, collaborative filtering techniques, including matrix factorization methods, achieve better recommendation results than content-based recommendation methods. Collaborative filtering technology uses a two-dimensiona...

AI Technical Summary

Problems solved by technology

However, there is no research based on interaction sequence data, which simultaneously includes three types of information: user's preference for items, dependencies between items and user's behavior similarity.

Images