[GRASS-stats] Spatial categories for risk data.

G. Allegri giohappy at gmail.com
Thu Dec 10 11:38:46 EST 2009


Hi Arthur, this is the classical task of clustering/classification
algorithms (like SVM, k-means, etc.)

For R take a look at the Cluster task page [1] for unsupervised
classification, and Machine Learning [2] for some supervised methods.

In Grass the classification/clustering methods are mostly directed to
imagery data, and their objective is to classify spectral images [3]
[4]. Anyway you could use it with other kind of imagery...

You also could take a look to this Python/C project, quite easy to
use, for supervised clasification: https://mlpy.fbk.eu/
And LIBSVM, which has various interfaces (Python, R, etc.):
http://www.csie.ntu.edu.tw/~cjlin/libsvm/

Hope it helps,
giovanni

[1] http://cran.r-project.org/web/views/Cluster.html
[2] http://cran.r-project.org/web/views/MachineLearning.html
[3] http://grass.osgeo.org/grass64/manuals/html64_user/i.cluster.html
[4] http://grass.osgeo.org/grass64/manuals/html64_user/i.maxlik.html

Il 10 dicembre 2009 17.08, HINTERMEISTER, ARTHUR W. (Artie)
<AHINTERMEISTER at amica.com> ha scritto:
> Hello,
>
>
>
> I currently have a map that looks something like this:
>
>
>
> -----------------------------------------
>
> |   x           x    o   x    o     x   |
>
> |  x     x         x       x     a o  a |
>
> |      a       x       o      o    x    |
>
> |  x      x       o      o     o     a  |
>
> |  x  x      x      o       a   o   a  o|
>
> |  x     x     o   x  o   x   a  a    a |
>
> | x   x    o      o     a   a      a    |
>
> -----------------------------------------
>
>
>
> x = low risk
>
> o = moderate risk
>
> a = high risk
>
>
>
> Each letter represents a person, and the entire map represents a state. Each
> person’s “risk score” is a continuous numeric value, but I have categorized
> it into low, moderate, and high for visual simplicity. What I’m hoping to do
> is draw lines that will divide out the categories. The lines do not have to
> necessarily be straight, but I would like them to be simple enough. I figure
> that I would have to account for any overlapping or outlying persons. This
> is my initial idea of how the map would turn out:
>
>
>
> -----------------------------------------
>
> |                  /             /      |
>
> |                 /             /       |
>
> |      x         /     o       /    a   |
>
> |               /             /         |
>
> |              /             /          |
>
> |             /             /           |
>
> |            /             /            |
>
> -----------------------------------------
>
>
>
> I have downloaded and installed GRASS and R, but I am not sure what type of
> spatial method or analysis to use. Could someone please offer me advice on
> how to get started with this project? Thanks so much!
>
>
>
>
>
> ------------------------------------------------------
>
> Arthur Hintermeister
>
> ahintermeister at amica.com
>
>
>
>
>
> _______________________________________________
> grass-stats mailing list
> grass-stats at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-stats
>
>


More information about the grass-stats mailing list