Talent recommendation method and device for scientific research knowledge graph based on graph neural network

Abstract

The invention discloses a method and device for recommending talents in a scientific research knowledge map based on a graph neural network, including: extracting the entity characteristics of each entity in the data of scientific research results papers to be processed and the association relationship information between entities, and establishing a scientific research knowledge map; Entity features constitute a unified feature representation of each node; through the unified feature representation and association relationship information, a graph neural network is constructed, and the graph neural network is trained to obtain the score value of each node; according to the score value of each author node, Get the prediction result of talent recommendation. The present invention enriches the information available in subsequent data mining by adding association relationships among various entities to generate different contribution weights, making the model more selective in the use of information, and taking the node in-degree value as the final adjustment An important numerical basis for the score value improves the model's ability to learn and predict.

Description

technical field [0001] The present invention relates to the field of machine learning talent recommendation algorithms, and more specifically, to a talent recommendation method and device based on a graph neural network for scientific research knowledge graphs. Background technique [0002] The recommendation and training of talents is an extremely important part of the development of scientific research. Using the talent recommendation algorithm to analyze the data of scientific research papers can help scientific research institutions recommend outstanding talents in the discipline, and provide reference opinions for talent introduction and training. There are many traditional talent recommendation algorithms, some of which are based on bibliometric methods to count the data related to the citations of papers and then recommend scholars with higher rankings. Using information such as the order of the papers' signatures and co-authorship relationships, and ignoring the dif...

AI Technical Summary

Problems solved by technology

This method achieves the effect of expert recommendation to a certain extent, but it is only limited to blog text features in social network data, ignoring the relationship features between users, and lacks the use of blog entity attribute features, such as the number of comments , the number of likes, etc., resulting in the inability to achieve a better expert recommendation effect

Images