GEMvis: a visual analysis method for the comparison and refinement of graph embedding models

Graph embedding, which constructs vector representation of nodes in a network, has shown effectiveness many graph analysis tasks, such as node classification, clustering, and link prediction. However, due to the complexity embedding models (GEMs) their nontransparency hyperparameters, evaluation comparison results retaining original features, consequently, selection suitable GEMs according tasks are challenging for people. In this paper, we present visual method, GEMvis, support from graph, metric, result spaces. The method also supports online refining GEM by tuning parameters its three components (graph sampling neural network structure, loss function). A series metrics, R_node measuring GEMs’ ability preserve specific R_degree R_closeness, is proposed quantitative features. Finally, case studies expert feedback illustrate GEMvis.