Soft document clustering using a novel graph covering approach.

Area of Science:

• Computer Science
• Information Science
• Data Mining

Background:

• Document clustering is essential for organizing and retrieving unstructured data in various scientific fields.
• Current methods often focus on hard clustering, limiting flexibility in data organization.

Purpose of the Study:

• To present a novel graph-theoretical approach for document clustering.
• To generalize graph partitioning to pseudostable sets and pseudocliques for enhanced clustering capabilities.
• To enable both soft and hard clustering for greater analytical flexibility.

Main Methods:

• Developed a graph-theoretical framework for document clustering.
• Generalized graph partitioning concepts to pseudostable sets and pseudocliques.
• Implemented integer linear programming and greedy approaches for problem-solving.

Main Results:

• The novel approach, PS-Document Clustering, demonstrates effectiveness on real-world datasets.
• The method successfully performs both soft and hard clustering.
• Valuable results were obtained on random and real-world data across various similarity measures.

Conclusions:

• PS-Document Clustering is a significant advancement in document clustering.
• The approach effectively integrates graph theory tools into document organization.
• The software is publicly available on GitHub for broader research application.