Weighted k means clustering

Hey all,

I am using Matlab for a geostatistical project.

I use coordinates of renewable energy facilities and try to optimize the electricity grid by clustering the facilities and finding the coordinates of some new electricity substations (cluster centroids).

I used to do that using k means algorithm.

Now I need to take into account the coordinates of the existing electricity substations and I need to use them with some weight, so the new substations will get closer to the old ones.

Does anybody know a way to use weight in k means? I found f kmeans algorithm, but I think it doesn't work really the way I need it to work.

Any ideas?

[Royi Avital]

I tried it myself as well. It seems something doesn't work there.

Answer 1

[Royi Avital]

Usually the weighting would be using Mahalanobis Distance Matrix.

If I'm correct about the file you linked, it uses a distance matrix which is Diagonal.

The Diagonal is determined by the weight vector.

[Royi Avital]

I tried it myself as well. It seems something doesn't work there.