You use 5 clusters because you have 5 classes of features???

Yes, Henning , k is an input of this algorithm, I want just to know how to specify the value of k ? most of the time is set empirically. I want just to know if there is a method or some rules that able us to fix it. 

Thank you for your reply Hamiderza :) 

A fine start point to solve your problem could be this wikipedia page: http://en.wikipedia.org/wiki/Determining_the_number_of_clusters_in_a_data_set

Thank you vincent 

"Automatically determining the number of clusters has been one

of the most difficult problems in data clustering. Most methods for

automatically determining the number of clusters cast it into the

problem of model selection. Usually, clustering algorithms are

run with different values of K; the best value of K is then chosen

based on a predefined criterion. ..."

Read the rest from Anil K. Jain's seminal work "Data clustering: 50 years beyond K-means". Good luck

Dear Khadidja

I suggest you dynamic or automatic clustering methods to extract the correct number of clusters

like ISODATA method but it requires two thresholds :one for splitting large clusters and one for merging two clusters 

you can also try metaheuristics (like GA, PSO DE) for finding correct number of clusters

Metaheuristic Clustering

Authors: Das, Swagatam, Abraham, Ajith, Konar, Amit

Good luck

Thank you for your reply :)