Empirical evaluation of constraint-based and score-based causal discovery algorithms

This thesis provides guidance on choosing appropriate algorithms for causal discovery by conducting comparative experiments on constraint-based and score-based methods. Causal discovery is an essential field in studying cause and effect, and the specific causal discovery problem under consideration in this thesis focuses on revealing underlying causal relationships from purely observational data, with sufficient assumptions to achieve this. Specifically, several algorithm variants based on two classes of causal discovery methods are evaluated using simulated datasets, with these datasets varied in characteristics including linearity of the dominating functions, sample size, and noise level. Results show the difference in numerical outcomes of different algorithm variants from the aspects of accuracy and execution time, leading to conclusions about the expected performances under known dataset characteristics. These conclusions are of high guiding significance when applying these algorithms to real-world datasets.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Han, Zifei

Contributors dc:contributor

• Kamalabadi, Farzad

• Causal Discovery
• Constraint-based Methods
• Score-based Methods

Statement dc:rights

• Copyright 2025 Zifei Han

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/129990

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F129990