The latest 2022 review of "graph classification research"

Graph data widely exists in the real world , Complex associations between composite objects and their elements can be naturally represented . The classification of graph data is a A very important and challenging issue , In Biology / There are many key fields of Applied Informatics , Such as molecular attribute judgment , New drug discovery, etc . But at the moment, There is still a lack of a complete overview of graph classification . Firstly, the definition of graph classification problem and the challenges in this field are given ; Then it combs and analyzes the classification methods of two kinds of graphs Law — Graph classification method based on graph similarity calculation and graph classification method based on graph neural network ; Then the evaluation index of graph classification method is given 、 Commonly used Comparison of data set and experimental results ; Finally, the common practical application scenarios of graph classification are introduced , The future research direction in the field of graph classification is prospected, and the full text is introduced Line summary .

Thesis link ：

http://www.jos.org.cn/jos/article/abstract/6323

This article No 1 This section gives the definition of graph classification problem and points out the problems and challenges in the field of graph classification . The first 2 This section combs the graph classification based on similarity calculation Method , It includes graph classification based on graph kernel method and graph classification based on graph matching . The first 3 This section introduces and analyzes the graph classification method based on graph neural network Law . The first 4 This section focuses on the evaluation of graph classification methods , Dataset including graph classification , Comparative analysis of the effect of evaluation index and some typical methods . The first 5 Section summary The application scenarios of graph classification in various fields are given, and the possible research trends in the future are given . The last section summarizes the full text .