Personalized recommendation method based on graph auto-encoder

Abstract

The invention discloses a personalized recommendation method based on a graph auto-encoder, and the method comprises the steps: building an adjacent matrix through the interaction between a user and an object, normalizing the adjacent matrix, carrying out convolution operation through a graph convolution network, and obtaining the hidden layer representation of each node; obtaining an initial feature vector of each node by taking a user comment text and an article description text as sources of node information, aggregating neighbor node features by using a graph attention network, and updating the node information; constructing a full-connection network by utilizing the attribute features of the user and articles to calculate to obtain hidden layer features; and splicing the hidden layerfeatures to obtain new node information, constructing a full-connection network, performing encoding, reconstructing scores of the user for the articles by using a bilinear decoder to serve as prediction scores, and generating a recommended article list by adopting TopN recommendation for the obtained prediction scores. Preference degree of the users to the articles can be analyzed more accurately, and the attention points of the users are found, so that more effective recommendation is carried out.

Description

technical field [0001] The invention relates to the technical field of text classification, deep learning, and recommendation system research, and in particular to a personalized recommendation method based on a graph autoencoder. Background technique [0002] With the continuous development of Internet technology, network information has grown explosively, but the rapid growth has also brought about the problem of information overload. Therefore, the importance of the process of accurately finding out what users are really interested in from a large amount of information is self-evident, and the proposal of the personalized recommendation algorithm is used to solve this problem, and it has gradually become an important aspect in the development prospect of the Internet. A big hot spot. [0003] Traditional recommendation methods such as collaborative filtering recommendation technology generally use user rating data on items to obtain user preferences. Although the algorit...

AI Technical Summary

Problems solved by technology

[0006] (2) The same rating given by the user cannot identify the user's concerns well

For example, when users give 5 points to the same item, some users focus on quality, while others focus on price. It is difficult for general recommendation systems to identify this difference.

Images