Knowledge social recommendation method, system and equipment based on graph neural network
A neural network and recommendation method technology, applied in biological neural network models, neural architectures, instruments, etc., can solve the problem of sparse scoring data recommending new user items, avoid data sparsity and cold start, avoid static and independent performance, improve the recommendation performance
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0048] For ease of understanding, see Figure 1 to Figure 3 , the present application provides an embodiment of a knowledge social recommendation method based on a graph neural network, including:
[0049] Step 101, build user item bipartite graph, extract user vector and item entity vector by graph self-encoder, described graph self-encoder adopts PinSage algorithm to aggregate nodes.
[0050] Such as figure 1As shown, in the embodiment of this application, firstly construct a user-item bipartite graph, and use the PinSage algorithm to aggregate nodes to construct a graph autoencoder, which can bring controllable number of neighbor nodes and can be based on neighbor nodes in the process of aggregating neighbor nodes. Importance aggregation, extracting user vectors and item entity vectors. The information transfer from item j to user i is defined as: c ij is the regularization constant, W r is the parameter matrix based on the scoring level, is the feature vector of it...
Embodiment 2
[0085] see Figure 4 , the application provides an embodiment of a graph neural network-based knowledge social recommendation system, including:
[0086] The bipartite graph unit is used to construct a user-item bipartite graph, extract user vectors and item entity vectors through a graph autoencoder, and the graph autoencoder uses the PinSage algorithm to aggregate nodes.
[0087] The user interest extraction unit is used to input the user vector into the RNN network to obtain the user interest vector output by the RNN network.
[0088] The network node update unit is used to put the collected user friend interest vectors and user interest vectors into the social network graph, and use the graph attention mechanism to update the nodes of the social network graph to obtain newer interest vectors.
[0089] The splicing unit is used to splice the update interest vector and the user interest vector to obtain the user's final embedding vector.
[0090] The entity field represent...
Embodiment 3
[0129] This application provides an embodiment of a knowledge social recommendation device based on a graph neural network. The device includes a processor and a memory:
[0130] the memory is used to store the program code and transmit the program code to the processor;
[0131] The processor is used to execute the knowledge social recommendation method based on the graph neural network in Embodiment 1 according to the instructions in the program code.
[0132] The device provided in the embodiment of this application mines the influencing factors of users and items from social networks and knowledge graphs respectively, calculates the embedding vectors of users and items, remodels and decodes, and fuses knowledge graphs and social networks to build recommendation models, avoiding The data sparsity and cold start encountered in single information recommendation solve the problem of the existing recommendation system’s rating data sparseness of users and items and the cold sta...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com