An Approach to Attribute Reasoning and Product Recommendation Based on Adaptive Graph Convolutional Networks

Abstract

The invention discloses an attribute reasoning and product recommendation method based on an adaptive graph convolutional network, including: 1. Constructing heterogeneous data: user rating matrix for products, user attribute matrix, product attribute matrix, user attribute index matrix, Product attribute index matrix; 2. Perform missing value filling preprocessing on user attribute matrix and product attribute matrix; 3. Obtain synergy matrix through one-hot encoding; 4. Construct feature fusion layer according to attribute matrix and synergy matrix; 6. Construct a prediction layer for attribute reasoning and product recommendation; 7. Update the node attribute matrix according to the output of the prediction layer; 8. Repeat steps 4 to 7 until the effect of attribute reasoning and product recommendation is optimal. The invention can fully exploit the high-order structural information of the graph, the internal interaction between node attributes, and the potential association between node attributes and link relationships, thereby realizing more accurate attribute reasoning and product recommendation.

Description

technical field [0001] The present invention relates to the field of attribute reasoning and product recommendation, in particular to an attribute reasoning and product recommendation method based on an adaptive graph convolutional network. Background technique [0002] The recommendation system effectively alleviates the problem of information overload, and has been successfully applied to platforms such as e-commerce, music, movies, and social networking. The model based on collaborative filtering is one of the most mainstream recommendation systems, which mines the user's historical records to make personalized product recommendations. The recommendation system based on collaborative filtering is widely used, but its performance is often limited by the sparsity of user behavior data. [0003] Attribute-enhanced collaborative filtering system uses rich attribute features of users (gender, age, occupation, etc.) and products (price, quality, category) to model users and pr...

AI Technical Summary

Problems solved by technology

Existing attribute reasoning works use semi-supervised graph learning (label propagation, graph regularization, deep models, etc.) to predict missing attributes, but the accuracy of attribute reasoning needs to be improved

Images