Clustering Algorithms

Does anyone know of a clustering algorithm which can cluster places with similar x and y co-ordinates into x clusters each with say 15 elements?

I've tried using k-means clustering but the only problem is you have to specify the number of clusters which is ok, but some of the clusters can have only 1 element and some can have 30, when for example i want 15 in each.

Any ideas?

That's not clustering, that's gerrymandering . Just kidding, and thinking about it.

How about this:

Set a number for your maximum range, I'll use 100 in this example

for each point, find the distance from it to each other point inside it's range (100) and subtract the distance from the max range (ie: dist 20=80, 1=99) Sum up all the converted distances, and this becomes the point's weight.

To find your clusters, you can either start at the fringes (the points with the smallest weights) or the centers (the points with the largest weights) of your clusters, and working off the nearest points, building up your clusters.

But it really depends on what your incomming data looks like, and how the clusters are going to be used. What if you have 30 points that are dense in the center, then spread out thinner at the edges? Do you want to split this area in half, or include them all into the same cluster, or maybe have the inside, denser area one cluster and the fringes in another? It really depends on the application.