Hello,

In my experience, I have found useful clustering methods that automatically determine the optimal number of clusters, without necessarily fixing it (like in k-means). For example, an interesting approach has been developed with a method called "affinity-propagation", which is based on passing messages between data points, and whose article I am referencing here.  There is also a R package called "apcluster" which compute this clustering, for those interested (also referenced below). 

But your question is how to fix the number of clusters, which I don't quite understand because the great majority of methods are based on fixing this number. Like I said, K-means is based on this K parameter which is the apriori-known number of clusters. Can you please give more information about what program or method are you trying to use that do not let you fix this number?

Well.. here the references to the articles I mentioned:

Article Clustering by Passing Messages Between Data Points

Article APCluster: an R package for affinity propagation clustering

I hope that helps. Best regards.

Maybe this post will help you

https://www.researchgate.net/post/How_to_find_number_of_cluster_in_cluster_dendrogram

to determine the best number of clusters, you can combine the following two things: First, suppose you make 9 times the clustering with the number of clusters 2,3,4 to 10 clusters, then graph the total distance per each cluster, choose the number of clusters that experienced the most significant change in total distance secondly, also consider what research objectives ... does the number of clusters based on the first consideration meet the research objectives, combine considerations 1 and 2 to get fixed number of clusters

Thanks for your valuable suggestions...

Dear Ajaya Karkee,

Hi,

CANOCO, NTSYS-pc, PC-ord are user-friendly software and propose the best possible situation; then, you can choose the best state. Also, you can try on R vegan package.

Regards,

Saeed