|
Activity Number:
|
659
- Recent Advances in Dimension Reduction and Clustering
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
|
Abstract #307132
|
|
|
Title:
|
Cluster Analysis via Random Partition Distributions
|
|
Author(s):
|
David Dahl* and Brandon Carter
|
|
Companies:
|
Brigham Young University and Brigham Young University
|
|
Keywords:
|
Hierarchical clustering;
Ewens-Pitman attraction distribution;
Chinese restaurant process
|
|
Abstract:
|
Cluster analysis is often used for exploratory data analysis in a variety of fields. Hierarchical clustering specifically is a heuristic method that uses as its input the pairwise distance among the items being clustered. Alternatively, we propose a new method for cluster analysis based on random partition distributions, such as the Ewens-Pitman Attraction (EPA) distribution. The EPA distribution is a probability distribution over all possible clusterings of the observations and is based on the same pairwise distance information used in hierarchical clustering. We seek to characterize the differences and similarities between hierarchical clustering and our clustering analysis based on random partition distributions. The advantages of this new method are illustrated through several case studies.
|
Authors who are presenting talks have a * after their name.
