Characterizing Class Differences in Attributed Graphs
Aria Rezaei, Bryan Perozzi, Leman Akoglu

Overview

AMEN Characterization 4 steps.

We find differences in classes of communities. This means that given two groups of closely tied subgraphs, we find the attributes that characterize them the most. These attributes can be used to reason on the differences in behavior among the two groups.

Slides

Code

You can find the most up-to-date version of the code on GitHub. Needless to say, this is a very research-y code, so be cautious while using it.

Dataset

Go to each dataset's webpage to see details and get instructions on how to use them.

  • Congress Co-sponsorship data. [download][info]
    • NOTE: A combination of two datasets here and here. Make sure to cite them properly.
  • DBLP Co-authorship data [download] [info]
  • Amazon Co-purchaseship data [download] [info]
    • NOTE: This dataset is a cleaned MATLAB version of this. Make sure to cite them properly.

Cite

If you find our work useful in your research, we ask that you cite the following paper:

@inproceedings{Rezaei:2017:TBC:3041021.3055138,
    author = {Rezaei, Aria and Perozzi, Bryan and Akoglu, Leman},
    title = {Ties That Bind: Characterizing Classes by Attributes and Social Ties},
    booktitle = {Proceedings of the 26th International Conference on World Wide Web Companion},
    series = {WWW '17 Companion},
    year = {2017},
    isbn = {978-1-4503-4914-7},
    location = {Perth, Australia},
    pages = {973--981},
    numpages = {9},
    url = {https://doi.org/10.1145/3041021.3055138},
    doi = {10.1145/3041021.3055138},
    acmid = {3055138},
    publisher = {International World Wide Web Conferences Steering Committee},
    address = {Republic and Canton of Geneva, Switzerland},
    keywords = {attributed graphs, community understanding, homophily, social networks, subspace discovery},
}