Repository for Benchmarking Causal Discovery Algorithms on Medical Data. Class project for CS520 Causal Inference and Learning @ UIC

Causal discovery algorithms infer the causal structure employing only observational data. Causal discovery is necessary to derive the causal model needed for performing causal inference and answering causal queries. Understanding and analyzing the causal relationships that exist among different features is crucial in several fields and applications like in the medical domain. In this paper, we perform the discovery and validation of causal relationships using medical observational data. State-of-the-art algorithms and conventional causal discovery techniques are compared using several evaluation metrics using synthetically generated data

• Wisconsin breast Cancer Dataset from the UCI repository
• Thyroid dataset from the UCI repository
• Cardiovascular Disease Dataset from kaggle
• Synthetic Dataset

• PC, Peter Clark
• GES, Greedy Equivalence Search
• LiNGAM
• NOTEARS
• SAM
• CGNN

Code implementation is in the code folder for all the algorithms.

• Causal discovery Toolbox
• Causalnex