How can I do constrained kmeans in Matlab?

I have an n*p matrix. I want to do normal kmeans but there is a limit such that the max number of sum of values in third column of the cluster should be 100. Can you please help

[Shubham Rawat]

Hi Ninto,

As per my understanding, you are trying to apply some condition on the cluster matrix. Could you please shed some more light on your question?

[Ninto Antony]

I have 600 points with me in a matrix with three columns. The first and second columns represent the latitude and longitude of the point The third column is a number(population) for that point. I want to cluster in such a way that the sum of all the population in one cluster is in between 300 and 350. If the above condition is met I want to get the centroid of that cluster.

Can you please guide me as to how I should proceed

Your help is much appreciated

Answer 1

[Shubham Rawat]

Hi Ninto,

As per my knowledge there is no as such function in MATLAB which can do constrained K-means clustering. In standard K-means, clustering is based upon only distances. You may have to build your own code for adding any constraints.

May be a good start would be if you:

• Assigned population to clusters.
• Then compute centroids of clusters.
• Assign points to centroids such that sum of population in that is in between 300-350.

[Ninto Antony]

So what you are saying is I do normal kmeans, take the centroids, then assign the points to the centroids I got earlier?

[Shubham Rawat]

You have to build your own code like K-means from scratch. In that you have to add constraints in the clusters that sum of the popluation is in between 300-350.

[Ninto Antony]

I can't even understand the code for normal kmeans. I think I would need some external help. But thanks for the solution 😊