NEW FEATURE:: Any thought on k-mean clustering function?

[rbucker]

I posted a question on the SQLite forums and was pointed here. There is something about the HISTO method that might be useful but looking at the results from my data... I'm not sure.

When doing a product search on Amazon there is filter helper... choose the price range. Clearly if you know enough about the products and their prices than pre-selecting the different ranges it's easy. But could the ranges be computed from a reasonable set of data.

The strategy that was recommended was k-mean clustering. After reading the problem description I did develop a solution. The results varied, reasonably, depending on selecting the right random set of centroids.

What do you think?

[Oeffner]

Hi Richard,
Implementing a k-mean clustering algorithm falls outside the scope of this project. But if you or someone else makes a pull request with a working piece of code I'll be happy to include it.

Having said that I think it would be relatively easy to let the user explicitly set the upper and lower borders of each bin in the histogram. This could give the user some control over the numbers of elements in each bin. I may get around to do this at some point.