Community search

Discovering communities in a network, known as community detection/discovery, is a fundamental problem in network science, which attracted much attention in the past several decades. In recent years, with the tremendous studies on big data, there is another related but different problem, called community search, which aims to find the most likely community that contains the query node, has attracted great attention from both academic and industry areas. It is a query-dependent variant of the community detection problem. In recent years, there has been many interestings studies focusing on this novel research problem.[1][2][3][4][5][6][7][8][9]

Main advantages

As pointed by the first work on community search[1] published in SIGKDD'2010, many existing community detection/discovery methods consider the static community detection problem, where the graph needs to be partitioned a-priori with no reference to query nodes. While community search often focuses the most-likely communitie containing the query vertex. The main advantages of community search over community detection/discovery are listed as below:

(1) More personalization.[2][8] Community detection/discovery often uses the same global criterion to decide whether a subgraph qualifies as a community. In other words, the criterion is fixed and predetermined. But in reality, communities for different vertices may have very different characteristics. Moreover, community search allows the query users to specify more personalized query conditions. In addition, the personalized query conditions enable the communities to be interpreted easily.

For example, a recent work,[8] which focuses on attributed graphs, where nodes are often associated with some attributes like keyword, and tries to find the communities, called attributed communities, which exhibit both strong structure and keyword cohesiveness. The query users are allowed to specify a query node and some other query conditions: (1) a value, k, the minimum degree for the expected communities; and (2) a set of keywords, which control the semantic of the expected communities. The communities returned can be easily interpreted by the keywords shared by all the community members. More details can be fround from.[9]

(2) More efficient. With the striking booming of social networks in recent years, there are many real big graphs. For example, the numbers of users in Facebook and Twitter are often billions-scale. As community detection/discovery often finds all the communities from an entire social network, this can be very costly and also time-consuming. In contrast, community search often works on a sub-graph, which is much efficient. Moreover, detecting all the communities from an entire social network is often unnecessary. For real applications like recommendation and social media markets, people often focus on some communities that they are really interested in, rather than all the communities.

Some recent studies[3][8] have shown that, for million-scale graphs, community search often takes less than 1 second to find a well-defined community, which is generally much faster than many existing community detection/discovery methods. This also implies that, community search is more suitable for finding communities from big graphs.

(3) Ease support for dynamically evolving graphs.[2] Almost all the graphs in real life are often evolving over time.  Since community detection often uses the same global criterion to find communities, they are not very sensitive of the updates of nodes and edges in graphs. On the contrary, community search can handle this easily, as it usually works on a sub-graph.

Measures for community search

Community search often uses some well-defined, fundamental measures of graphs. The commonly used measures are minimum degree,[1][3][5][6][8] k-truss,[4][7] k-edge-connneted, etc. Note that, the minimum degree measure is also used for defining the k-core of a graph. Among these measures, the minimum degree measure is the most popular one, and has been used in many recent studies.

References

  1. 1 2 3 Mauro Sozio and Aristides Gionis. 2010. The community-search problem and how to plan a successful cocktail party. In Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining (KDD '10). ACM, New York, NY, USA, 939-948. DOI=http://dx.doi.org/10.1145/1835804.1835923
  2. 1 2 3 Wanyun Cui, Yanghua Xiao, Haixun Wang, Yiqi Lu, and Wei Wang. 2013. Online search of overlapping communities. In Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data (SIGMOD '13). ACM, New York, NY, USA, 277-288. DOI=http://dx.doi.org/10.1145/2463676.2463722
  3. 1 2 3 Wanyun Cui, Yanghua Xiao, Haixun Wang, and Wei Wang. 2014. Local search of communities in large graphs. In Proceedings of the 2014 ACM SIGMOD International Conference on Management of Data (SIGMOD '14). ACM, New York, NY, USA, 991-1002. DOI=http://dx.doi.org/10.1145/2588555.2612179
  4. 1 2 Xin Huang, Hong Cheng, Lu Qin, Wentao Tian, and Jeffrey Xu Yu. 2014. Querying k-truss community in large and dynamic graphs. In Proceedings of the 2014 ACM SIGMOD International Conference on Management of Data (SIGMOD '14). ACM, New York, NY, USA, 1311-1322. DOI=http://dx.doi.org/10.1145/2588555.2610495
  5. 1 2 Rong-Hua Li, Lu Qin, Jeffrey Xu Yu, and Rui Mao. 2015. Influential community search in large networks. Proc. VLDB Endow. 8, 5 (January 2015), 509-520. DOI=http://dx.doi.org/10.14778/2735479.2735484
  6. 1 2 Nicola Barbieri, Francesco Bonchi, Edoardo Galimberti, and Francesco Gullo. 2015. Efficient and effective community search. Data Min. Knowl. Discov. 29, 5 (September 2015), 1406-1433. DOI=http://dx.doi.org/10.1007/s10618-015-0422-1
  7. 1 2 Xin Huang, Laks V. S. Lakshmanan, Jeffrey Xu Yu, and Hong Cheng. 2015. Approximate closest community search in networks. Proc. VLDB Endow. 9, 4 (December 2015), 276-287. DOI=http://dx.doi.org/10.14778/2856318.2856323
  8. 1 2 3 4 5 Yixiang Fang, Reynold Cheng, Siqiang Luo, Jiafeng Hu. 2016. Effective community search for large attributed graphs. Proc. VLDB Endow. 9, 12, 1233-1244.
  9. 1 2 http://i.cs.hku.hk/~yxfang/acq.html
This article is issued from Wikipedia - version of the 10/11/2016. The text is available under the Creative Commons Attribution/Share Alike but additional terms may apply for the media files.