Research on the Nodes Importance in Complex Networks Based on Community Partition

F Chen, Y Huang - Proceedings of the 2nd International Conference on …, 2019 - dl.acm.org
F Chen, Y Huang
Proceedings of the 2nd International Conference on Big Data Technologies, 2019dl.acm.org
In complex network analysis, the importance of nodes has always been one of the main
research fields of graph theory and the most common calculation method is the PageRank
algorithm. The existing research is mainly based on the whole network level to rank all
nodes, which will bring a huge amount of computation when the network is quite large. In
fact, a node is more likely to be affected by the nodes that are closely connected to it, rather
than by the nodes that have the highest importance in the entire network. So, it is necessary …
In complex network analysis, the importance of nodes has always been one of the main research fields of graph theory and the most common calculation method is the PageRank algorithm. The existing research is mainly based on the whole network level to rank all nodes, which will bring a huge amount of computation when the network is quite large. In fact, a node is more likely to be affected by the nodes that are closely connected to it, rather than by the nodes that have the highest importance in the entire network. So, it is necessary to decompose the networks into communities, whose main feature is that the nodes within a community are closely connected, while the nodes between different communities are sparsely connected. In this paper, the classical Louvain algorithm will be applied to achieve community partition, for the further study of nodes importance. Firstly, the statistical characteristics of PageRank values of nodes between different communities are studied for four different types of networks: regular network, random network, small-world network and scale-free network. Secondly, a real enterprise investment network is studied based on a large number of Chinese enterprise investment data. The nodes in the same community are extracted to form a sub-network and a new set of PageRank values will be calculated for these nodes. Comparison results of two sets of PageRank values of the same group of nodes show that the rank of nodes importance in the whole network is highly consistent with that in the sub-network. For the network model established by big data, community partition method can greatly reduce the computational load of nodes important and simplify the research process. The conclusion of this paper is applicable to large-scale complex networks in reality, and has vital practical significance.
ACM Digital Library
Showing the best result for this search. See all results