Information recommendation method for neural collaborative filtering based on graph convolution

Abstract

The invention provides an information recommendation method for neural collaborative filtering based on graph convolution. The method comprises the following steps: S1, collecting user behavior data and attribute contents of users and articles; S2, if the collected behaviors are explicit scores, creating a user-user graph according to score information of the users on the articles, and calculating the similarity between the users and the similarity between the articles to construct a user-user graph and an article-article graph; if the collected behavior information is only implicit interaction information such as browsing and clicking, constructing a user-article graph; S3, performing graph convolution operation on the constructed relation graph to obtain feature vectors of nodes of users and articles; S4, fully connecting the feature vectors of the nodes of the users and the articles with the attribute features of the users and the articles; and S5, taking the obtained feature vectors of the nodes of the users and the articles as an input layer of a neural collaborative filtering algorithm framework, performing prediction, and performing information recommendation according to a prediction result.

Description

Technical field[0001]The present invention relates to the field of information recommendation techniques, and more particularly to an information recommendation method based on the chassis volume of neuro-filtered.Background technique[0002]According to the 45th "China Internet Development Status Statistical Report", as of January 2020, my country's netizens has reached 904 million, and the Internet has reached 64.5%, and the user size and usage rate of all kinds of Internet applications are continuous. Sexual growth mode. As the Internet gradually integrates people's daily life, traditional search engines have not been able to meet people's needs, in order to quickly and accurately predict users' preferences, the recommended system plays a very important role to help users can find in massive data. I like items. However, the recommended system is still facing some problems.[0003](1) The recommendation of the recommended system is mainly based on the collection of users' score data, ...

AI Technical Summary

Problems solved by technology

[0003] (1) At this stage, the recommendations of the recommendation system are mainly based on the collection of user rating data. However, some users are unwilling to leave ratings because they are unwilling to disclose their privacy or waste their time. sparsity of

[0004] (2) The traditional collaborative filtering algorithm based on matrix decomposition uses a simple inner product method to calculate complex user and item features in a low-dimensional space, and cannot deeply learn the relationship between user and item features

[0005] (3) The traditional recommendation method does not give too much consideration to the attributes of users and items, as well as the interaction relationship between users and users, and the interaction relationship between items and items.

Images