Seminar on Computational Learning and
Adaptation
Data Mining Based on Database Set Operations and Rough Set Theory
Xiaohua Tony Hu
Department of Math and Computer Science, San Jose State University
tonyhu@mathcs.sjsu.edu
In this talk I will discuss a new data mining approach based on database set operations and rough set theory. I will first give an introduction of rough set theory, and then show how to modify and redefine the key concepts of rough set theory (reduct, core, superfluous attributes) in the context of database set operations in order to improve the computational efficiency of rough sets for data mining applications. Next I will discuss a novel context-sensitive measure of merit for features, the rough set based feature selection method and a rule induction algorithm that can be used to generate a globally optimal set rules from the data. I will also present the prototype data mining system DBROUGH-II and some of its example applications. I conclude by presenting a novel approach to constructing an ensemble of classifiers that improve learning accuracy.
Date: Thursday, February 14
|
Time: 4:15-5:30PM
|
Place: Cordura 100
|
Return to the seminar schedule