Loading...

Community Detection in Social Networks using Node Attributes and Topology of the Network

Radmanesh, Mohammad Reza | 2015

876 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 47655 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza
  7. Abstract:
  8. In the last decade, participation of hundreds of millions of users in social networks, has turned social networks into a powerful tool to steer the flow of information. As a result, the study of different aspects of these networks has become one of the indispensable interests of many researchers. One of the most important research problems in social network analysis is the community detection problem. From structural point of view, a community is a set of nodes with more internal links than external links, and from behavioral point of view, a community is a set of nodes with similar attributes and features. Most of previous works on community detection have focused on the structure of the network, while there are other sources of information such as the attributes of nodes. Recently, researchers have developed some algorithms for community detection, which consider both aspects (structure of the network, and attributes of the nodes) simultaneously. These algorithms can be divided in to two categories and the main challenge of them is how to merge different sources of information. The first category is Model based Algorithms which propose a generative model for the network based on a stochastic process. The results of Model based Algorithms mostly depends on the descriptive model. The second category is Non-model based Algorithms, which try to improve some measures of community structure in the network. Although the precision of Non-model based Algorithms is lower than the first category, but they take the advantage of having less complexity which makes them more applicable for real large networks.In this research we proposed an algorithm for community detection using the structure of network and the attributes of nodes, which is categorized in Non-model based methods.The proposed algorithm uses a biased random walk with restart to detect nodes of each community. The main contribution of the algorithm is considering different impacts for node attributes in community detection. In other words, the algorithms learns impact of each attribute per community. We have evaluated the functionality of our proposed method on real data. Experimental results show that the proposed algorithm works better than other recent methods in community detection which utilize both sources of information
  9. Keywords:
  10. Social Networks ; Network Structural Features ; Community Detection ; Communities in Network ; Network Nods Attribute

 Digital Object List