Kim Keen Wee
will defend her thesis

REGION DISCOVERY USING SUPERVISED CLUSTERING
 

Abstract

The goal of data mining is the discovery of valid, novel, interesting and potentially useful patterns in datasets. The discovery of interesting regions in spatial datasets is becoming an important data mining task. In this thesis, we investigate the use of supervised clustering for region discovery. Supervised clustering deviates from traditional clustering in that it is applied on classified examples with the objective of identifying clusters with a high density with respect to a single class. Three different measures of interestingness that serve as the fitness function for supervised clustering algorithms are proposed. Two different supervised clustering algorithms, SRIDHCR and SCEC, are modified for the purpose of region discovery.  Techniques for the visualization of the results of region discovery are proposed. Finally, we discuss the results of experiments that apply and evaluate the developed region discovery techniques on a benchmark consisting of artificial and U.S. census-based spatial datasets.

 

Date: Wednesday, May 4, 2005
Time: 3:00 PM
Place: 550-PGH



Faculty, students, and the general public are invited.
Thesis Advisor: Dr. Christoph Eick