Many approaches have been proposed to find a subgraph that contains the query node, and the goodness metric is maximized or minimized, such as k-core [6, 9, 10], k-truss [11, 12] and densest graph .
In this paper, instead of introducing a new goodness metric for community searches like k-core  or k-truss , we consider the problem of a community search from a new point view: local distance dynamics.
There are many different definitions of cohesive subgraphs in the literature, including maximal clique , k-core , k-truss , and maximal k-edge connected subgraph .
In recent years, several metric-based community search models have been studied, such as k-core , k-truss , and k-influential community .
The k-core and k-truss algorithms belong to the first class and are required to optimize both internal denseness and external sparseness.
It can be seen that the K-Truss method is better than other methods on most networks.
(2) For K-Truss, we find that K-Truss has better results and stability than the K-Core algorithm on real-world networks.
It is important to note that the K-Core runs faster than K-Truss and K-Hop, but its accuracy is low.
The K-Truss method has high relative density and slightly outperforms other algorithms.
Comparing three algorithms, we can find that K-Truss and K-Hop have the better value and stability.
The K-Core and K-Truss are very close, and the accuracy of K-Truss is next.
The K-Truss algorithm also uses a local search procedure but has to compute the support of each edge on the whole graph; thus, it has a much longer running time than K-Core.