Statistical Analysis of Design Aspects on Various Graph Embedding Learning Classifiers

Graphs representation learning is an emerging research topic. Graph representation learning generates vectors that capture immense graphs' structure and properties. The quality of graph representation vectors affects machine learning downward tasks like node categorization, edge estimation and anomaly detection. For efficient graph representation vectorization, both traditional graph embedding techniques and methods based on graph neural networks (GNN) have been described. These approaches work on “static & dynamic graphs”. In contrast to dynamic graphs which constantly adds new nodes and edges, static graphs consist of a single unchanging structure. This paper reviews traditional and GNN-based embedding approaches for both static and dynamic graphs. This paper summarizes GNN challenges and proposes a framework. Finally, real world applications of these techniques are explored.