Fig 1.
Node 1 shares 4 neighbouring nodes (squares) with node 2. Since this is an undirected network, the MNS of node 1 to node 2 is 4, and vice versa.
Fig 2.
Nodes A to G are divided into 3 groups (diamond, square and circle). The numbers attached to the edges are the MNS of a pair of nodes. Numbers in red indicates the highest MNS with another node. It must be noted that not all possible related nodes and edges are depicted in this figure.
Table 1.
Example of the capacity of a given node for accepting a neighbouring node into its community.
Fig 3.
The flowchart of CLPA-GNR.
Table 2.
Summary of the generated LFR networks.
Table 3.
Summary of generated GN networks.
Table 4.
Summary of the real-world networks.
Fig 4.
The NMI comparison on undirected and unweighted small LFR benchmark networks with small average degree.
The size of the networks, size of the communities and the average degree are n, C and k, respectively.
Fig 5.
The NMI comparison on undirected and unweighted small LFR benchmark networks with large average degree.
The legend is the same as in Fig 4.
Fig 6.
The NMI comparison on undirected and unweighted big LFR benchmark networks with small average degree.
The legend is the same as in Fig 4.
Fig 7.
The NVI comparison on undirected and unweighted small LFR benchmark networks with large average degree.
Fig 8.
The NVI comparison on undirected and unweighted big LFR benchmark networks with large average degree.
The legend is the same as in Fig 7.
Fig 9.
The NVI comparison on undirected and unweighted GN and RC benchmark networks. The size of community for RC ranges from 2 to 157.
Table 5.
The values of μ and D when CLPA-GNR, LPA, Infomap, GANXiS and NIBLPA obtain trivial detection.
Table 6.
The NVI of real-world networks with ground truth communities.
Bold values are the best detection algorithm for each network.
Table 7.
The modularity (Q) for all the real-world networks.
Bold values are the best detection algorithm for each network.
Table 8.
The modularity density (Qds) for all the real-world networks.
Bold values are the best detection algorithm for each network.