[R-sig-Geo] complete linkage Agglomerative hierarchical clustering, nnclust, spatclus or something else?

[Roger Bivand]

On Tue, 20 Apr 2010, Hans Ekbrand wrote:

> I have just read about clustering on wikipedia, and learnt that what I
> want is:
>
> Agglomerative hierarchical clustering, with complete linkage

library(cluster)
?hclust

is for clustering with moderate numbers of points.

Below: in UTM, the units are metres, so distances are in metres too.

Hope this helps,

Roger

>
> I searched for suitable r-packages for this and found nnclust, and
> spatclus. Are those the packages that you could recommend for
> clustering events data (the events here is urban fires, created by
> arsonists)? Or do you want to suggest other packages?
>
> In this first analysis I want to do the clustering should only by
> location, and ignore the point in time.
>
> Later on I will also include time data, so if the clustering package
> could handle time too, that would be great, but that is not a
> requirement at this time.
>
> The aim at this stage is only to group events with the same, or almost
> the same, location. In my data-set the coordiante-data is a bit too
> precise in some cases.
>
> I guess a crude way of clustering would be to round the
> coordinate-data to a lesser number of significant digits, however, a
> sound clustering algorithm would be better :-)
>
> I also need to grasp the scale here:
>
> With this proj4string, and datapoints as below:
>
>  ..@ proj4string:Formal class 'CRS' [package "sp"] with 1 slots
>  .. .. ..@ projargs: chr " +proj=utm +zone=33 +ellps=GRS80 +units=m +no_defs"
>
> head(fires at coords, 2)
>         East   North
> [1,] 315359.9 6393110
> [2,] 325862.4 6405239
>
> dist(head(fires at coords, 2))
>         1
> 2 16044.10
>
> Is this in meters?
>
> TIA
>
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no