Re: K-Means Clustering

[John Darrington]

Hi Mehmet!

As I mentioned, the CLUSTER command is soemthing which I think it would be
great to support.

One issue with clustering is its memory complexity. It requires O(n^2) where
n is the number of cases being clustered.
Have you tested your algorithm with large numbers of cases?

Maybe Ben has some ideas how an efficient distance matrix can be  implemented
in PSPP (maybe sparse-array.c can help?) .

In any case, I'd be interested to see your code, and the results of your 
comparisons.  Can you post them somewhere?


J'

On Tue, Mar 08, 2011 at 05:56:37AM -0800, Mehmet Hakan Satman wrote:
     hi everybody,
     
     I am interested in PSPP and i read about something about the needs for 
     developing some functionality.
     I implemented a k-means clustering  library using the GNU scientific 
library and 
     sent an informative e-mail to John. He suggested me to join this group and 
share 
     my ideas with the stuff. 
     
     
     I compared the results with SPSS outputs. The analysis of variance table 
is not 
     completed but we may add this feature.
     
     I would be glad to integrate something to PSPP and work with you.
     
     What do you think about this?
     
     
           
-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

signature.asc
Description: Digital signature