A systematic analysis and guidelines of graph neural networks for practical applications
作者：
Highlights：
• We analyze the state-of-the-art methods of graph neural networks (GNN) in four phases.
• 1300 combinations of methods are compared with 13 well-known benchmark datasets.
• It results in five guidelines to use GNN for practical graph-related problems.
• Comparative experiments with more than 3600 runs support the analysis and guidelines.
• Experimental reproducibility and replicability are verified by comparing with the literature.
摘要
•We analyze the state-of-the-art methods of graph neural networks (GNN) in four phases.•1300 combinations of methods are compared with 13 well-known benchmark datasets.•It results in five guidelines to use GNN for practical graph-related problems.•Comparative experiments with more than 3600 runs support the analysis and guidelines.•Experimental reproducibility and replicability are verified by comparing with the literature.
论文关键词：Graph neural network,Graph embedding,Deep learning,Graph classification,Social network analysis,Bioinformatics
论文评审过程：Received 29 September 2020, Revised 15 May 2021, Accepted 21 June 2021, Available online 25 June 2021, Version of Record 29 June 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115466
