[postgis-users] Combining / agglomerating / clustering numbers of adjacent polygons

[Derek Jones]

Dear Matias

Thanks for the thought. Any help is always appreciated.


I actually am using this agglomeration (when I eventually derive it  ;-) 
) to seed kmedoids or kmean centroids for statistical clustering, (doing 
it in R via a variety of methods to see what works best). I already have 
a much larger set of points that need to be clustered around the set of 
centroids I generate from this grid agglomeration process.

However doing this gridding and agglomeration process first gives me a 
twofold advantage, in that it not only gives me the centroid for the 
clustering of the "other" data set, but also allows me to identify the 
actual spatial areas in which some of my (statistical) cluster points 
(in the other data set) should fall to start with, so I will do at least 
one level of spatial intersection, then one level of statistical 
clustering on top.


I am at this point favoring doing the grid agglomeration 
programmatically - perhaps I can get that done today. (Unless anyone 
comes up with an SQL solution before I finish coding it up ;-))

I'll let you folks know if I make the programmatic approach work but I'm 
still open to other input in the meantime.

Thanks to all for suggestions thus far.

Kind regards

Derek.