Activity Number:
|
588
- Statistical Learning: Clustering
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
Abstract #322419
|
View Presentation
|
Title:
|
Topological Probabilistic Classification (TopProC)
|
Author(s):
|
Fairul Mohd-Zaid* and Christine Schubert Kabban
|
Companies:
|
Air Force Research Lab and Air Force Institute of Technology
|
Keywords:
|
Topology ;
L-moment ;
Clustering ;
Bootstrap
|
Abstract:
|
We propose a novel classification and clustering method by bootstrapping topological structures within a dataset and discriminating the classes within the dataset using the L-moments of the degree distribution from the bootstrapped structures. Topological Data Analysis is a growing field, and although it provides a new perspective for discovering non-linear patterns and relationships within the dataset, it is still a manual and visual process which requires some intuition from the user. TopProC combines the clustering and topological approach of TDA with computational statistics to build the distributions of the L-moments from the degree distribution of the topological structure by treating it as a graph. The L-moment distributions can characterize any group of interest based on a given set of labels which can then be used for classification. The method also allows the distribution of measures from the dataset to be built for each cluster within the topological structure which summarizes the features of said cluster and can be used to explain the driving factor for each cluster and subsequently the group of interest. Preliminary result is demonstrated on several applied datasets.
|
Authors who are presenting talks have a * after their name.