Graph Convolutional Networks: An Analysis of Method and Applications in Different Fields and Systems

Abstract

With the prosperity of the information era and the maturity of computer technology, artificial intelligence has attracted much more attention. The emergence of Graph Neural Networks (GNNs) as a theory that can process non-Euclidean structure data such as graph data makes more applications possible. Graph Convolutional Neural Networks (GCNs), as a theoretical branch of GNNs, uses new theories to innovate and optimize on the basis of inheriting the ideas of predecessors, allowing the rapid development of this field. In this article, the author mainly introduces the basic theory of converting graph data into Euclidean structure data, which is the most important part of GCNs, distinguishing from Convolutional Neural Networks (CNNs). The successful applications of GCNs in the fields of recommendation systems and traffic prediction are also listed. Through the analysis of theory and applications, the shortcomings and development prospects of GNNs are discussed. Finally, the author points out that GCNs still have room for improvement in terms of data scale, network layers, and dynamics and complex nature of graph data.