clustring with the same number of point inside each cluster

I have x and y coordinate of a set of points resulting in matrix X. As I know, idx = kmeans(X,k) is designed in a way that I can fix the number of clusters to k. However, I want to fix an additional parameter too. I want to fix the number of points inside each cluster too.

Let me give a simple example. Assume we have 99 points (and thier x and y coordinate in matrix X). Also, we know that there should be 11 clusters (that is k) and each one of clusters should contains 9 points inside. Do you have any ideas how I can force MATLAB to put exactly 9 points inside each cluster? It would be nice if it can be also shown as plot or gscatter. Thanks

[KSSV]

Share the data....can give a try.

[Mehdi Foumani]

For example, assume the following code. Is it possible to have 11 clusters there (each with exactly 9 points inside)?

n=99;

x = randi([0 1000],n,1);

y = randi([0 1000],n,1);

all_xy=[x,y];

[idx,C] = kmeans(all_xy,11);

[~,idx] = pdist2(C,all_xy,'euclidean','Smallest',1);

gscatter(all_xy(:,1),all_xy(:,2),idx)

for i=1:n

text(all_xy(i,1),all_xy(i,2),num2str(idx(i)));hold on;

end

Answer 1

[Abhishek Kumar]

Dear User,

As per my understanding you are able to use the k-means function in MATLAB, but you seek a functionality to fix number of points in each cluster as well. I would like to inform that such a functionality is not available right now with k-means functions. We, have recorded your issue and such a functionality might be available in upcoming releases.

[David Morales]

I hope it gets implemented soon :D