Attribute and Struct-Aware Node Mining and Visual Exploration

Graph node retrieval is a crucial application, such as identifying compounds with similar properties in a protein network. Traditional methods rely on similarity metrics from graph topology, like graph kernel methods, but neglect vital node attribute information. As network size grows, these methods become computationally costly. To address this, we propose an innovative node retrieval method that seamlessly integrates node topology and attributes using deep learning. This approach is effective in complex system networks, as demonstrated by our developed graph exploration visualization system, validated through a user study confirming its practicality and effectiveness.