GAM: Cluster hunting software (v4.0)
Centre for Computational Geography
University of Leeds
The Cluster Hunter program gives you several ways to analyse clusters in your
data. This webpage will take you through one - analysing data with GAM. It should
be noted that Cluster Hunter is an incomplete experimental program and behaves as such.
The first thing to do is get the data file you want to use. To follow the
information on this page you need a comma separated
text file with data in the following
ID number, Eastings, Northings, Incidences, Total population in ID area.
The program assumes the ID number refers to an area with a total population,
in which there are a number of incidences of something. The Eastings and
Northings refer to the centre of the area. Note that the ID cannot include
Once you have the file, open up Cluster.jar
Either double-click on the file in Windows Explorer, or if this doesn't work, open
a DOS command prompt, find the file and type...
> d:\jre1.5\bin\java -jar cluster.jar
You'll have to replace the "d:\jre1.5\bin\" bit with a reference to wherever the
JRE is on your machine.
(For those used to Java, the file is a
The application is composed of the following...
- A File menu, for opening files and quiting.
- A Select menu, for selecting the cluster hunting method.
- A Parameters menu, for selecting parameters for the various methods.
- A Run menu, for running once you've opened the file and selected a method.
- A Display menu, for displaying your data, the results, and overlaying boundaries on both
from a shapefile.
- A message space below the menus.
The first thing to do is to open up the file.
Pick Load Data File from the File menu, and open your
file (note that the file must have the .dat extension). This may take a
few minutes. Cluster should tell
you that the file has been opened and the number of points that have been
loaded into the application from the file when it's finished.
Once you have the files in, you can look at the data points using the
Display Database item on the Display menu. You may have to
adjust the size of the window that pops up in order to see the data properly.
To close the window (indeed any of the windows) go to the File menu
and select Quit (Note that you can't use the "X" in the corner of windows
to close them).
Next you have to pick a clustering method.
Choose GAM from the Select menu. If you go back to
the menu you'll see it's now got a little star next to it.
The other options are
Knn a type of K-means detection, and Random, which
finds clusters like GAM, but by randomly throwing circles at the problem (it's
used to test the effectiveness of algorithms against a benchmark of random guessing).
You'll see that the Set Parameters item on the Parameters menu
can now be selected. This would let you set the parameters. We're
just going to run GAM with the defaults.
Next, run GAM
Pick the Go option from the Run menu. GAM will
run and various messages will appear in the messages area. It make take
several minutes for GAM to run. When it's finished, the message area should
Finally we want to display and save the results.
Pick Display Results from the Display menu.
This will bring up a window showing the clusters. You can use the
Load Shapefile item on the Geography menu of this
window to load a boundary dataset over the results to see where the clusters
are (this is a little flaky with large boundary datasets).
Once you're happy with the results you can use Save as Jpeg item
on the File menu to save the clusters as an image. However, these are pretty poor
quality and it
won't presently save your boundaries, so it may be better to save an image of
your clusters by taking a screenshot (Prt Sc button on your
keyboard, then open up Microsoft Photo Editor and select Paste as New Image
from the Edit menu).
Alternatively you can export the cluster file as a Arc GRID compatible file. This is an
ASCII file that is suitable for conversion to a raster/GRID format using ArcToolbox.
To shut down Cluster, go to the main window, and select
Quit from the File menu.
[School of Geography homepage]