/eggs

AAAI '20 STARAI Workshop paper: "EGGS: A Flexible Approach to Relational Modeling of Social Network Spam."

Primary LanguagePython

logo

Binder

EGGS (Extended Group-based Graphical-models for Spam) is an approach that combines multiple relational-modeling techniques, specifically stacked graphical learning (SGL) and probabilistic graphical modeling (PGM), on top of a traditional classifier; the result is a comprehensive model that can flexibly capture any independent and/or relational signal in the data.

Install

  1. Clone or fork this repository.
  2. Install dependencies:
    • Install Python >= 3.5.
      • Install packages using requirements.txt.
    • Probabilistic Soft Logic LINK
      • Requires Java 8.
      • Install the package pslpython.
    • Libra LINK

Paper

@article{brophy2020eggs,
  title={Eggs: A flexible approach to relational modeling of social network spam},
  author={Brophy, Jonathan and Lowd, Daniel},
  journal={arXiv preprint arXiv:2001.04909},
  year={2020}
}