.png){kind=icon}
.png){kind=icon}
Return
GrAR: A novel framework for Graph Alignment based on Relativity concept
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2021.115908.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Social media will continue growing rapidly and integration of social media information has become important. Information integration and many tasks require graph alignment or finding the same nodes in different networks. There are several methods for graph alignment. Many of these methods introduce a new feature vector to represents node structure. Then compare all node pairs between two graphs without using any graph information. According to our current knowledge, many methods use either local node properties or global node properties, and fewer methods use node's structural role for graph alignment. We propose a relativity based graph alignment framework (GrAR) that takes a feature vector for each node or ID vector as input and extracts structural role properties for each node for graph alignment task. Experimental results on different datasets indicate that if the ID vector is strong enough, the proposed framework improves the efficacy of the ID vector. Furthermore, to deal with the presence of noise, a relative-degree feature based on the concept of relativity is introduced. Our results show that the proposed feature vector is noise-resistant and increases the efficiency of graph alignment tasks alone or within the framework.
Keywords:
GrAR
Relativity concept
Graph alignment
Social Network Analysis
Anchor Link
Entity Resolution
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W