A method of cluster analysis
Loading...
Authors
Cima, William Michael
Subjects
Advisors
Date of Issue
1970-06
Date
Publisher
Language
en_US
Abstract
A method of cluster analysis is presented in which
points in n-dimensional space are analyzed through a subdivisive
procedure. The points are orthogonally projected
onto that line which maximizes their variance and the
resulting point distribution is then analyzed with the use
of a histogram. Wherever possible, divisions between conglomerates
of points are made and each separate clump is
subsequently analyzed. Ultimately adjacent groups are
combined and analyzed through an analogous technique in an
effort to re-unite any points which may have inadvertently
deviated from the group with which they truly associate.
The method is later refined to allow the detection of groups
in several point dispersions which would have appeared as a
single conglomeration under the original method. An example
is given to illustrate the applicability of the procedure.
Type
Thesis
Description
Series/Report No
Department
Operations Analysis
Organization
Naval Postgraduate School