[postgis-users] Re: Cluster Analysis

[Eli Dylan Lorimer]

Hi Paul,
I think I'll try and go your suggested PL/R route. Is it true that I  
can essentially send the results of a spatial query to R and invoke  
*any* R function available on said data? I'm going to look into this  
anyway and hopefully get it all to gel well with GeoTools such that I  
can continue to make my requests over WFS.

As to your question about clusters on the fly, I like your idea of  
having the static table that I update via trigger, but I really need  
to dynamically pass a few parameters which will change the resulting  
clusters. Hum, I'll put some more thought into this idea though as  
that would be so much faster to do.

Cheers,
-dylan


Dylan,

Are you calculating clusters on the fly? Do you need to? How about
having a clusters table, that is hooked to your base data table with
a trigger? That will be just as up-to-date as your base data, and
only have to do the calculations once per data update.

Paul

PS. A distance threshold "cluster" will return lots of unexpected and
unwanted results, which is why more sophisticated tricks have been
developed.