A framework for similarity-based clustering of biomedical citation graphs

In many scientific fields, identifying groups of similar items within complex systems is crucial, often referred to as community detection in networks. This research focuses on community detection in citation graphs of PubMed documents. While existing methods have explored clustering citation graphs, they often neglect article metadata or rely on non-scalable solutions. In this thesis, I propose a framework for community detection in any citation graph within the PubMed database, utilizing both article metadata and citation graph structure. The core of this framework is a similarity metric that measures document relevance by incorporating metadata and citation importance weights. Scalability is achieved through a parallel framework for data gathering and feature construction, demonstrated by clustering a 7.5 million node graph. The approach is flexible, allowing users to adjust feature importance and content. The effectiveness of this method was evaluated using the Leiden CPM clustering algorithm across various feature combinations to assess their impact on cluster quality. Additionally, the approach’s potential for application beyond citation graphs, such as in social networks, is discussed. In summary, the developed similarity metric enhances community detection by integrating both topology and metadata, providing a foundation for future research in this area.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Mohasel Arjomandi, Hossein

Contributors dc:contributor

• Chacko, George

• Clustering
• Scientometrics
• Community Detection
• Biomedical Documents

Statement dc:rights

• Copyright 2024 Hossein Mohasel Arjomandi

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/127150

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F127150