Limits...
Identifying influential nodes in large-scale directed networks: the role of clustering.

Chen DB, Gao H, Lü L, Zhou T - PLoS ONE (2013)

Bottom Line: Inspired by these facts, we propose a local ranking algorithm named ClusterRank, which takes into account not only the number of neighbors and the neighbors' influences, but also the clustering coefficient.Subject to the susceptible-infected-recovered (SIR) spreading model with constant infectivity, experimental results on two directed networks, a social network extracted from delicious.com and a large-scale short-message communication network, demonstrate that the ClusterRank outperforms some benchmark algorithms such as PageRank and LeaderRank.Furthermore, ClusterRank can also be applied to undirected networks where the superiority of ClusterRank is significant compared with degree centrality and k-core decomposition.

View Article: PubMed Central - PubMed

Affiliation: Web Sciences Center, School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, People's Republic of China ; Department of Physics, University of Fribourg, Fribourg, Switzerland.

ABSTRACT
Identifying influential nodes in very large-scale directed networks is a big challenge relevant to disparate applications, such as accelerating information propagation, controlling rumors and diseases, designing search engines, and understanding hierarchical organization of social and biological networks. Known methods range from node centralities, such as degree, closeness and betweenness, to diffusion-based processes, like PageRank and LeaderRank. Some of these methods already take into account the influences of a node's neighbors but do not directly make use of the interactions among it's neighbors. Local clustering is known to have negative impacts on the information spreading. We further show empirically that it also plays a negative role in generating local connections. Inspired by these facts, we propose a local ranking algorithm named ClusterRank, which takes into account not only the number of neighbors and the neighbors' influences, but also the clustering coefficient. Subject to the susceptible-infected-recovered (SIR) spreading model with constant infectivity, experimental results on two directed networks, a social network extracted from delicious.com and a large-scale short-message communication network, demonstrate that the ClusterRank outperforms some benchmark algorithms such as PageRank and LeaderRank. Furthermore, ClusterRank can also be applied to undirected networks where the superiority of ClusterRank is significant compared with degree centrality and k-core decomposition. In addition, ClusterRank, only making use of local information, is much more efficient than global methods: It takes only 191 seconds for a network with about [Formula: see text] nodes, more than 15 times faster than PageRank.

Show MeSH
The ratio of the number of final recovered nodes by ClusterRank to those by out-degree centrality, PageRank and LeaderRank.The non-overlapped nodes in the top-50 lists are initially infected. We set . Each data point is obtained by averaging over 100 independent runs.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3814409&req=5

pone-0077455-g007: The ratio of the number of final recovered nodes by ClusterRank to those by out-degree centrality, PageRank and LeaderRank.The non-overlapped nodes in the top-50 lists are initially infected. We set . Each data point is obtained by averaging over 100 independent runs.

Mentions: Figure 6 shows resulted from the top-50 most influential nodes at different infected rates . It can be seen that resulted from the top-50 most influential nodes by ClusterRank is larger than that by other ranking algorithms. Figure 7 shows the ratio of the number of ever infected (i.e., finally recovered) nodes resulted from top-ranked nodes by ClusterRank to those by other ranking algorithms at different infected rates . Note that, in figure 7, only non-overlapped node appeared in the top-50 lists by ClusterRank and other ranking algorithms are initially set to be infected. The ratio is up to 2 when for Delicious network (see figure 7(a)) and it approaches 20 (surprisingly high) when for SM network (see figure 7(b)). In fact, some nodes in the SM network are of very large out-degree but the out-degree of their followers are very small. These nodes are not as important as their out-degrees indicate, and ClusterRank could dig out really influential nodes and assign the high-degree-yet-low-influence nodes low ranks.


Identifying influential nodes in large-scale directed networks: the role of clustering.

Chen DB, Gao H, Lü L, Zhou T - PLoS ONE (2013)

The ratio of the number of final recovered nodes by ClusterRank to those by out-degree centrality, PageRank and LeaderRank.The non-overlapped nodes in the top-50 lists are initially infected. We set . Each data point is obtained by averaging over 100 independent runs.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3814409&req=5

pone-0077455-g007: The ratio of the number of final recovered nodes by ClusterRank to those by out-degree centrality, PageRank and LeaderRank.The non-overlapped nodes in the top-50 lists are initially infected. We set . Each data point is obtained by averaging over 100 independent runs.
Mentions: Figure 6 shows resulted from the top-50 most influential nodes at different infected rates . It can be seen that resulted from the top-50 most influential nodes by ClusterRank is larger than that by other ranking algorithms. Figure 7 shows the ratio of the number of ever infected (i.e., finally recovered) nodes resulted from top-ranked nodes by ClusterRank to those by other ranking algorithms at different infected rates . Note that, in figure 7, only non-overlapped node appeared in the top-50 lists by ClusterRank and other ranking algorithms are initially set to be infected. The ratio is up to 2 when for Delicious network (see figure 7(a)) and it approaches 20 (surprisingly high) when for SM network (see figure 7(b)). In fact, some nodes in the SM network are of very large out-degree but the out-degree of their followers are very small. These nodes are not as important as their out-degrees indicate, and ClusterRank could dig out really influential nodes and assign the high-degree-yet-low-influence nodes low ranks.

Bottom Line: Inspired by these facts, we propose a local ranking algorithm named ClusterRank, which takes into account not only the number of neighbors and the neighbors' influences, but also the clustering coefficient.Subject to the susceptible-infected-recovered (SIR) spreading model with constant infectivity, experimental results on two directed networks, a social network extracted from delicious.com and a large-scale short-message communication network, demonstrate that the ClusterRank outperforms some benchmark algorithms such as PageRank and LeaderRank.Furthermore, ClusterRank can also be applied to undirected networks where the superiority of ClusterRank is significant compared with degree centrality and k-core decomposition.

View Article: PubMed Central - PubMed

Affiliation: Web Sciences Center, School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, People's Republic of China ; Department of Physics, University of Fribourg, Fribourg, Switzerland.

ABSTRACT
Identifying influential nodes in very large-scale directed networks is a big challenge relevant to disparate applications, such as accelerating information propagation, controlling rumors and diseases, designing search engines, and understanding hierarchical organization of social and biological networks. Known methods range from node centralities, such as degree, closeness and betweenness, to diffusion-based processes, like PageRank and LeaderRank. Some of these methods already take into account the influences of a node's neighbors but do not directly make use of the interactions among it's neighbors. Local clustering is known to have negative impacts on the information spreading. We further show empirically that it also plays a negative role in generating local connections. Inspired by these facts, we propose a local ranking algorithm named ClusterRank, which takes into account not only the number of neighbors and the neighbors' influences, but also the clustering coefficient. Subject to the susceptible-infected-recovered (SIR) spreading model with constant infectivity, experimental results on two directed networks, a social network extracted from delicious.com and a large-scale short-message communication network, demonstrate that the ClusterRank outperforms some benchmark algorithms such as PageRank and LeaderRank. Furthermore, ClusterRank can also be applied to undirected networks where the superiority of ClusterRank is significant compared with degree centrality and k-core decomposition. In addition, ClusterRank, only making use of local information, is much more efficient than global methods: It takes only 191 seconds for a network with about [Formula: see text] nodes, more than 15 times faster than PageRank.

Show MeSH