Construction and Application of Knowledge Graph based on Graph Neural Networks

Abstract

In recent years, as an emerging graph data learning technology, Graph Neural Network (GNN) has received very extensive attention. Graph Neural Network (GNN) mainly provides Graph Embedding, a technology that can be used for graph representation learning. The introduction of traditional graph analysis expands the processing capabilities of deep learning for non-Euclidean data. Provides a method to extract features from irregular data. It is widely used in social networks, recommendation systems, financial risk control, physical systems, molecular chemistry, life sciences, knowledge graph, traffic forecasting and other fields. This article first introduces the basic construction methods and simple applications of the knowledge graph. For example: using neo4j to build a small movie knowledge graph, you can easily query the director and actors corresponding to the movie. And on the basis of constructing a knowledge graph, it has realized the detection of fraud rings, searched for account holders who shared multiple legal information, inquired about their possible financial risks, and sorted them by risk. Secondly, the Random Walk, Node2vec, GCN and GraphSAGE are introduced around graph representation learning. Finally, based on the knowledge graph and graph neural network, the company's untrustworthy risk is predicted based on the data. Relying on the powerful data processing capabilities of the relational data of the graph neural network, an accuracy rate of more than 95% is obtained.