[R-sig-Geo] Natural Breaks - Jenks

[Roger Bivand]

On Tue, 18 Apr 2006, Ben Brehmer wrote:

> I have implemented Jenks' algorithm (for finding the natural breaks) in 
> php thanks to some sample code I found at: 
> http://www.mail-archive.com/r-sig-geo@stat.math.ethz.ch/msg00290.html . 
> This algorithm is also implemented in ArcView to determine natural 
> breaks in the legends.
> Currently I am running the algorithm on a data set which has 65 000 
> elements in it, which takes over 3 hours (due to a nested for loop). 
> ArcViews' implementation on the other hand returns within seconds. Would 
> anyone possibly know why ArcViews implementation is so much more efficient.

library(classInt)
?classIntervals
y <- runif(65000)
yClass <- classIntervals(y, n=5, style="fisher")

runs on a 1.5GHz machine in 225 seconds. This is using the Fortran code 
you refer to directly. My guess is that Arc looks at the number of unique 
values, and, if there are many, uses a heuristic. If it sampled and set 
the seed the same each time, the result would be the same, and the code 
runs acceptably fast for say 2000 values. Maybe Arc also precomputes 
values?

Roger

> 
> Any help would be greatly appreciated.
> 
> Ben Brehmer
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

-- 
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