Complex network analysis of keywords co-occurrence in Graph Neural Networks

Systematic studies of scientific literature are essential for outlining the present state of research and highlighting additional growth channels in a field of study, but systematic studies are essentially tedious, time-consuming, and hand operated. Recently, keyword co-occurrence networks (KCNs) are utilized for knowledge mapping. In a KCN, each keyword is represented as a node, and each co-occurrence of a pair of words is represented as a link. In this work, we propose a KCN-based approach that can be implemented before undertaking a systematic review to guide and accelerate the review process. The novelty of this method lies in the new metrics used for statistical analysis of a KCN that differ from those typically used for KCN analysis. The approach is demonstrated through its application to Graph Neural Network literature. The KCN approach identified the knowledge components, knowledge structure, and research trends that match with those discovered through a traditional systematic review of the GNN field. Because KCN-based analyses can be conducted more quickly to explore a vast amount of literature, this method can provide a knowledge map and insights before undertaking a rigorous traditional systematic review. This two-step approach can significantly reduce the effort and time required for a traditional systematic literature review. The proposed KCN-based pre-systematic review method is universal. It can be applied to any scientific field of study to Graph neural network prepare a knowledge map.

کلیدواژه ها:

co-occurrence networks (KCNs) ، knowledge mapping ، Graph Neural Network literature ، Network analysis