Related to your query, I suggest you to follow https://au.mathworks.com/help/stats/classification-using-nearest-neighbors.html

Hi Anselme，

Do check the following links for explanations on how you may go about solving the problem.

https://www.analyticsvidhya.com/blog/2019/08/comprehensive-guide-k-means-clustering/

https://realpython.com/k-means-clustering-python/

Good luck!

please indicate what software you are using

some packages in R do that automatically

If you want to learn theory then go for Elbow method for optimal calue of K. Otherwise, it will be calculated automatically by packages available in R or Python.

Thanks all for your reacts. Your explanations are very interesting. Elbow seems to approximatively solve the problem of defining the best selection of K-value for a dataset, very powerful. In my case, I will try to evaluate the distance between the centroid and the last faraway point in each cluster and will try to consider that as the cluster rayon and check if that solves definitely the problem.

I think this chapter is useful for you ...

http://infolab.stanford.edu/~ullman/mmds/ch7.pdf