Relation and fuzzy clustering for document categorization and analysis

• Main Author: Mei, Jian-Ping
• Other Authors: Chen Lihui
• Format: Theses and Dissertations
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Pattern recognition; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/48627
• Institution: Nanyang Technological University

This thesis focuses on the investigations of using fuzzy clustering for automatic document categorization based on relations between document and other types of objects. Three approaches called Fk-Parts, LinkFCM and FC-MR are proposed to handle the document clustering problem under different scenarios. We start with a basic situation, and propose Fk-Parts to cluster documents based on document-document relation. The new mechanism of using multiple weighted medoids to represent each cluster makes Fk-Parts perform better than single medoid based approaches. After that, we consider situations where both vector representation of documents and document-document relation are available. LinkFCM is then formulated by incorporating relation into the well known fuzzy c-means approach, so that both types of data are considered in clustering. Finally we propose a fuzzy approach of multi-type relational data clustering FC-MR. This approach simultaneously clusters documents and other types of objects based on the relations among them.