Mining groups of common interest: Discovering topical communities with network flows

Liyun Li, Nasir Memon

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper tackles the problem of detecting topical communities from within an organization by mining readily available network access pattern information. A Bayesian generative process is used to model the behavior of user's network access pattern and thereby her consumption of online content. The idea is that users within same topical interest group tend to share similar online access patterns. By leveraging this pattern, along with side information of domain-names and keywords within the accessed websites, one is able to model these observations under the framework of a mixed membership statistical model. Hence the access patterns of users-to-websites, as measured at the edge of an organization's network boundary, can be decomposed into constituent topical communities without any human effort in selecting specific features. Experimental results on real-world network flow trace demonstrate that the proposed method can effectively detect topically meaningful community structures. Besides better detection accuracy of communities compared with other community detection methods, the proposed method can detect interesting but non-evident hidden communities which cannot readily be detected by other known methods.

Original languageEnglish (US)
Title of host publicationMachine Learning and Data Mining in Pattern Recognition - 9th International Conference, MLDM 2013, Proceedings
Pages405-420
Number of pages16
DOIs
StatePublished - Aug 13 2013
Event9th International Conference on International Conference on Machine Learning and Data Mining, MLDM 2013 - New York, NY, United States
Duration: Jul 19 2013Jul 25 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7988 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other9th International Conference on International Conference on Machine Learning and Data Mining, MLDM 2013
CountryUnited States
CityNew York, NY
Period7/19/137/25/13

    Fingerprint

Keywords

  • Generative Model
  • LDA
  • Network Flow
  • Topical Community Detection

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Li, L., & Memon, N. (2013). Mining groups of common interest: Discovering topical communities with network flows. In Machine Learning and Data Mining in Pattern Recognition - 9th International Conference, MLDM 2013, Proceedings (pp. 405-420). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7988 LNAI). https://doi.org/10.1007/978-3-642-39712-7_31