What is CLUSTER 3.0?
CLUSTER 3.0 performs hierarchical cluster analysis on datasets consisting
of up to 200 rows and 50 columns, with a choice of different clustering
algorithms and similarity coefficients. A particular feature of CLUSTER
is stratigraphically constrained analysis. In normal cluster analysis the
ordering of the objects in the data matrix is not involved. Especially
in Earth Sciences, the spatial ordering of objects (generally the vertical,
stratigraphical or layering order) is important. CLUSTER allows to keep
this order intact.
Input data files are: ASCII row/column files, DBASE or Paradox files.
The output consists of a tree diagram, that can be exported to an Enhanced
Windows Graphics metafile for further processing in generic graphics editors
such as Corel Draw WP Draw.
Operating system requirements: Windows 95/98/NT. The current version
is a Windows upgrade of a DOS program, originally written for my thesis
research. See that publication (Van Huissteden,
1990) for applications.
What is cluster analysis?
Cluster analysis assists classification of objects into groups of similar
objects. The method used in CLUSTER results in a hierarchical classification:
individuals are linked in groups, these groups are linked into larger groups.
The result is shown in a 'family tree' of objects. Cluster analysis consists
of two components: A 'distance metric' or similarity coefficient that measures
the similarity between objects and groups of objects; A clustering algorithm,
which determines how similar objects groups are linked into larger groups.
Different combinations of metric and algorithm are possible, with different
effects on the results. Metrics included in CLUSTER are: Euclidian distance,
chord distance, Manhattan metric. Algorithms: Minimum variance, centroid
sorting, nearest neighbour, furthest neighbour, weighted and unweighted
pairgroup methods.
License agreement.
The software is distributed as freeware, commercial reselling is not allowed.
Claims with regard to support or damage caused by improper or proper use
of this program cannot be accepted by the designer of this software, nor
his employer. If the program is successfully used in the preparation of
a scientific publication, it is to be mentioned in the acknowledgements.
Installation.
The program, its help file and a sample data file is stored in a self-extracting
zip file, CLUSTER30.EXE. Put in the directory where you want to install
it, doubleclick the file icon, and it will be unzipped. Further installation
is not required, although you may wish to make appropiate links in the
Windows Program menu or create a shortcut on the desktop. Please complete
the form below before downloading.
Bugs.
Cluster needs a database driver from Borland to open .DBF files. On computers where this driver is not installed, .DBF
files cannot be opened. In that case, convert your DBF files to ASCII text. I am looking for a way to solve this problem.
The Windows metafile of the tree diagram does not import nicely into Microsoft Word. This is not an error of Cluster, it is Bill Gates' problem.
New feature.
A 'Copy to Clipboard' button is added to the graphics window. It copies the tree diagram as a bitmap to the Windows clipboard.
Updated: December 11, 2001
Please complete the form below. This is just to keep track of who is
using the software and to keep you informed on changes in the software. It will not cause any
unnecessary mailing.
Download CLUSTER30
Homepage J. van
Huissteden