Exploiting Tie-strength and Structure Towards Link Prediction in Social Networks

No Thumbnail Available
Date
2017
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Analysis of complex network has emerged as a booming research area since last decade. Social network, a type of complex network, has gained attention from the contemporary researchers, due to the abundance of social network data in the Web in recent times. Rapid increase in the number of subscribers to the social platforms (such as blogs, dating sites, friends making sites) provided by the Web has revealed unseen human relationships, and motivated the researchers to make good use of this. This thesis deals with an important problem of social network analysis (also of complex network analysis): link prediction. Given a social network, the link prediction problem predicts new relationships which will appear in future. Homophily, i.e., similarity between two individuals influences new connections. This work models homophily by combining link strength and structure of the network towards link prediction. Link strength is encoded in link weight in several ways, which is derived from pattern of dyadic interaction between two nodes. Structural homophily is captured by traditional proximity based link prediction methods like common neighbor, Jaccard's coeffcient, Adamic/Adar etc.
Description
Supervisors: Sanasam Ranbir Singh & Sukumar Nandi
Keywords
COMPUTER SCIENCE AND ENGINEERING
Citation