TY - CHAP

T1 - The impact of partially missing communities on the reliability of centrality measures

AU - Martin, Christoph

N1 - Conference code: 6

PY - 2018

Y1 - 2018

N2 - Network data is usually not error-free, and the absence of some nodes is a very common type of measurement error. Studies have shown that the reliability of centrality measures is severely affected by missing nodes. This paper investigates the reliability of centrality measures when missing nodes are likely to belong to the same community. We study the behavior of five commonly used centrality measures in uniform and scale-free networks in various error scenarios. We find that centrality measures are generally more reliable when missing nodes are likely to belong to the same community than in cases in which nodes are missing uniformly at random. In scale-free networks, the betweenness centrality becomes, however, less reliable when missing nodes are more likely to belong to the same community. Moreover, centrality measures in scale-free networks are more reliable in networks with stronger community structure. In contrast, we do not observe this effect for uniform networks. Our observations suggest that the impact of missing nodes on the reliability of centrality measures might not be as severe as the literature suggests.

AB - Network data is usually not error-free, and the absence of some nodes is a very common type of measurement error. Studies have shown that the reliability of centrality measures is severely affected by missing nodes. This paper investigates the reliability of centrality measures when missing nodes are likely to belong to the same community. We study the behavior of five commonly used centrality measures in uniform and scale-free networks in various error scenarios. We find that centrality measures are generally more reliable when missing nodes are likely to belong to the same community than in cases in which nodes are missing uniformly at random. In scale-free networks, the betweenness centrality becomes, however, less reliable when missing nodes are more likely to belong to the same community. Moreover, centrality measures in scale-free networks are more reliable in networks with stronger community structure. In contrast, we do not observe this effect for uniform networks. Our observations suggest that the impact of missing nodes on the reliability of centrality measures might not be as severe as the literature suggests.

KW - Business informatics

UR - http://www.scopus.com/inward/record.url?scp=85036630496&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-72150-7_4

DO - 10.1007/978-3-319-72150-7_4

M3 - Article in conference proceedings

AN - SCOPUS:85036630496

SN - 978-3-319-72149-1

VL - 1

T3 - Studies in Computational Intelligence

SP - 41

EP - 52

BT - Complex Networks & Their Applications VI

PB - Springer

T2 - 6th International Conference on Complex Networks and Their Applications - Complex Networks 2017

Y2 - 29 November 2017 through 1 December 2017

ER -