topic Re: K-means clustering in Spatial Statistics Questions

K-means clustering

AstridHarendza — Wed, 12 Dec 2012 15:10:57 GMT

Hello,

I have a question regarding k-means clustering in ArcGIS. I have a shapefile, which contains a number of polygons with different values for mean, standard deviation, skewness and quantiles. I would like to cluster them using k-means clustering. I am aware that ArcGIS 10.1 just integrated a tool for this, but I am still working under ArcGIS 10.0. I also discovered a toolbox, which is using R in ArcGIS, but I'm having problems with the installation of the required Design package in R and hence can't get it to work.

Accordingly I was wondering if you could suggest a different way on how to get k-means clustering to work in ArcGIS 10.0. Can I do the clustering in R and then import the results somehow into ArcGIS? Tips and suggestions are highly appreciated!

Many thanks for your help,
Astrid

Re: K-means clustering

JeffreyEvans — Sat, 15 Dec 2012 03:36:16 GMT

Astrid,
See my post in this thread for related R code. Please note that I am using a K-Medoids algorithm to avoid some know problems with K-means. There is also code for selecting the optimal k based on the partition silhouettes.

http://forums.arcgis.com/threads/27922-R-Point-Clustering-Tools-Error?highlight=cluster

I was using an optional maptools syntax so please note that the last line of the code is incorrect and should be:
writeOGR(shape, getwd(), "cluster_results", driver="ESRI Shapefile")

Re: K-means clustering

AstridHarendza — Mon, 17 Dec 2012 07:44:24 GMT

Thank you very, very much Jeffrey! This solved my problem :)!

Re: K-means clustering

MuhammedThouseef_MT — Wed, 03 Dec 2014 16:52:52 GMT

Hi Jeffrey,

Let me know how to use K-Medoid algo in ArcGIS ?