Loading...

Prediction of The Link Sign Between Nodes in Signed Social Networks

Malekzadeh, Mohammad | 2010

383 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 42122 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Rabiee, Hamid Reza
  7. Abstract:
  8. Interactions in social networks consist of positive relations, such as friendship, trust, like, and negative relations, such as antagonism, distrust, dislike. “Signed networks” are utilized to model these networks. These networks are presented by “signed graphs” in which nodes are the people and relations are modeled by sign of edges. One of the challenging problem in signed networks is link sign prediction, i.e., specifying unknown edge sign along with evolution of the network given sign of some edges and further information about remainder of network. Two approaches are used to answer this problem. The first approach is proposing computing models for sign prediction. In this assertion we propose two frameworks based on two prominent sociopsychological theories, namely, balance and status. In these frameworks two new measures are introduced for signed networks analysis. First one is Pleasure measure, in analysis of undirected networks and the second one is rationality measure, in analysis of directed networks. Our implementations on real world datasets -Epinions, Slashdot and Wikipedia- demonstrate that in comparison with current models, these frameworks have accurate results . The second approach is proposing analyzing models to follow the evolution of these Networks. In this regard, based on pleasure measure and in tandem with network creation game, a model called best response model is developed. We will show that in this game calculating the best response for each player is NP-hard problem. Therefore, we will propose an approximate solution to the problem using quadratic programming. Furthermore, several theorems related to Nash equilibrium point of the game are presented. We will prove that the game graph will change to balanced graph after O(n) time step of the game. Having this model, popular social media sites are analyzed and new measure called smartness is employed in the process. Finally, we extend our best response model. The extended model fit more appropriately in evolution of social media sites.
  9. Keywords:
  10. Social Networks ; Game Theory ; Signed Networks ; Social Balance

 Digital Object List

  • پيش بيني علامت يال مابين گره ها در شبكه هاي
  •   view

 Bookmark

No TOC