AutoCD

Towards Automated Causal Discovery

This repository contains the code for the paper:
Towards Automated Causal Discovery: a case study on 5G telecommunication data
Konstantina Biza, Antonios Ntroumpogiannis, Sofia Triantafillou, Ioannis Tsamardinos
https://arxiv.org/pdf/2402.14481.pdf

Overview

AutoCD is a causal discovery framework that aims to fully automate the application of causal discovery.

It can be applied to a plethora of real-world problems with :

• cross-sectional or temporal data
• high-dimensional data
• unmeasured confounders
• mixed data types

AutoCD consists of three modules:

1. Automated Feature Selection (AFS)
   • reduces the dimensionality of the problem, by selecting a set of features that optimize a user-defined target
2. Causal Learning (CL)
   • learns a causal model over the selected features
3. Causal Reasoning and Visualization (CRV)
   • visualizes and interprets the learned causal model, as a response to a set of user-defined queries

Packages

AutoCD uses the following publicly avalaible implementations

• MXM R package: https://CRAN.R-project.org/package=MXM
  • feature selection
• Tetrad project: https://github.com/cmu-phil/tetrad
  • causal discovery on cross-sectional and temporal data
  • data simulation
• Tigramite project: https://github.com/jakobrunge/tigramite
  • causal discovery on temporal data
  • data simulation

It also needs the following python packages:

• scikit-learn
• pandas
• numpy
• py4cytoscape
• JPype1
• networkx

AutoCD visualizes the graphs using the Cytoscape platform: https://cytoscape.org/

Notes

You need to download R, Java and Cytoscape to run AutoCD.
Make sure that Python, R, Java and Cytoscape are installed in the same folder (e.g. Program Files)

Contact

[email protected]