An improved local bridging centrality model for distributed network analytics

Classifying an edge or a node within a network graph according to its bridging characteristics is an important structural measure that has many practical analytic applications. Bridging centrality is a relatively new graph-based centrality metric for classifying nodes serving as key structural connections between dense components. A global bridging centrality has been previously introduced and requires global knowledge of the entire graph structure to perform the centrality computation. Recently, a local bridging centrality variant was also introduced that can be calculated locally requiring only limited local neighborhood graph information. We introduce an extended model of the recently introduced local bridging centrality based upon the the concept of a local “friends of friends” or 2-hop neighbor distance egocentric graph. We further develop and analyze the use of this extended centrality with both unweighted and weighted graph structures. Finally we present a series of comparative studies using a variety of graph models and examine ranking correlation comparisons with global bridging centrality results. Our analysis includes comparisons with both past literature models and newer temporal graph results of a 100-node mobile wireless network scenario with link weights representing dynamic reception quality.