Ding-Zhu Du, Principal Investigator

Clustering Evaluation

This project includes three categories:

• Information retrieval
• Text document classification
• Optimal clustering

The group has developed a new weighting scheme for a vector space model to improve retrieval performance for an information-retrieval system. The main idea of this approach is to consider not only present terms but also absent terms in finding similarities among document vectors. Optimal clustering, in this research, will provide hierarchical clustering algorithms with a clustering criterion. This criterion is based on the centroid approach and the squared error method that have been the most widely-used evaluation measures for clustering techniques. The group’s approach will be demonstrated and verified using the uncertainty of clustering optimality. By applying this feasible clustering criterion to document classification, the group will obtain multiple optimal centroids, which will reduce classification errors.

Research Group and Collaborators

Yunjae Jung, Graduate Student Researcher
Haesun Park, Faculty Collaborator