Robust influence angle for clustering mixed data sets

The great importance in attempting to identify clusters of observations which may be present in a data is how close observations are to each other. Two observations are close when their dissimilarity is small. Some traditional distance functions cannot capture the pattern dissimilarity among the obs...

Full description

Saved in:
Bibliographic Details
Main Author: Aziz Nazrina
Format: Conference Proceeding
Language:English
Subjects:
Clustering
Covariance matrix
Robustness
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The great importance in attempting to identify clusters of observations which may be present in a data is how close observations are to each other. Two observations are close when their dissimilarity is small. Some traditional distance functions cannot capture the pattern dissimilarity among the observations. The other demand is the dissimilarity measurement should have the ability to deal with a variety of data types. This article proposed a new dissimilarity measure namely, Robust Influence Angle (RIA) based on eigenstructure of the covariance matrix and robust principal component score. The proposed measurement is able to identify cluster of observation and it also has the ability to handle data set with mixed variables.
ISSN:0094-243X
1551-7616
DOI:10.1063/1.4887699