Orange Forum • View topic - On orngClustering

On orngClustering

A place to ask questions about methods in Orange and how they are used and other general support.

On orngClustering

Postby srikanthvc » Fri May 21, 2010 13:18

While using the k-means algo, noticed teh following. Pl comment.

1. For unimodal distributions i.e. only single attribute, it does not work (So had to repeat that data in the next column as well to get results)

2. The silhoutte values, are always in descending order, the maxm being for 2. This is absed on a few tests, but it has always been like this.

How to go about deciding the no. of clusters then?

Thanks for the great work!

Return to Questions & Support