|
Activity Number:
|
343
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #309157 |
|
Title:
|
Methods for K-Models Clustering
|
|
Author(s):
|
Li Li*+ and James E. Gentle
|
|
Companies:
|
George Mason University and George Mason University
|
|
Address:
|
4400 University Dr, Fairfax, VA, 22030,
|
|
Keywords:
|
data mining ; clustering ; k-means ; simulation
|
|
Abstract:
|
The K-means method is often used in exploratory data analysis to partition data into k groups. The underlying idea is to form groups so as to minimize within-group variation. This is generally effective if the groups are roughly spherical and are separated from each other. If, however, the groups have other structures, such as linear patterns, k-means may not be effective in identifying the groups. Motivated by these considerations, we propose a "K-models" clustering method. We evaluate its performance and study the effect of different starting strategies for K-models clustering.
|
