Luni 3 decembrie 2007, cu începere de la ora 16:00, în amfiteatrul 2/I va avea loc următoarea conferinţă destinată studenţilor. Vor fi abordate şi subiecte legate de cooperări bilaterale în folosul studenţilor. Toţi cei interesaţi sunt invitaţi să participe.

Probabilistic Community Discovery for Large-Scale Social Networks

SPEAKER

John Yen and Hank Foley, College of Information Sciences and Technology, The Pennsylvania State University

Abstract

Large-scale social networks have been rapidly formed on the Web, due to the popularity of social network websites and the increasing amount of time people spent on them.

One of the challenges in analyzing these networks is to partition them based on the structure of the network so that interesting clusters (i.e., communities) can be identified automatically. In this talk, we describe two probabilistic approaches for community discovery based on Latent Dirichlet Allocation (LDA) model: one based on the conventional network encoding scheme (that considers only direct neighbors) and the other one that considers indirect neighbors. Using collaborative networks in two disciplines (computer science and nanotechnology), we compare these two approaches. The clusters generated are evaluated using perplexity analysis and compactness measures. The experiment suggested that an LDA-based community discovery model using a network encoding scheme that considers indirect neighbors is a promising approach for analyzing large-scale social networks.