Canonical Correlation Analysis of Neighborhood-based Centrality Metrics vs. Shortest Path-based Centrality Metrics
- Resource Type
- Conference
- Authors
- Meghanathan, Natarajan
- Source
- 2023 Congress in Computer Science, Computer Engineering, & Applied Computing (CSCE) CSCE Computer Science, Computer Engineering, & Applied Computing (CSCE), 2023 Congress in. :1433-1438 Jul, 2023
- Subject
- Computing and Processing
Measurement
Correlation
Network analyzers
Canonical correlation
Centrality metrics
Neighborhood-based metrics
Shortest path-based metrics
Complex network analysis
- Language
Canonical correlation analysis is useful to analyze the correlation between two sets of features in a dataset. In this paper, we demonstrate the use of canonical correlation analysis to study the correlation between the neighborhood-based centrality metrics (Degree centrality: DEG and Eigenvector centrality: EVC) vs. the shortest path-based centrality metrics (Betweenness centrality: BWC and closeness centrality: CLC) for a suite of 80 complex real-world networks. We observe more than 4/5th of the 80 real-world networks analyzed to exhibit either a strong positive correlation or negative correlation between these two sets of centrality metrics (i.e., if the neighborhood-based centrality value for a node is lower, the shortest path-based centrality metric value for the node is likely to be lower or higher). We also observe the extent of variation in node degree to play a significant role in the correlation between these two sets of centrality metrics in complex real-world networks.