Combine performance vectors for k-means clustering

zebra

Hi,

I want to combine two PerformanceVectors (centroid distances from "Performance (Cluster Distance Performance)" and the cluster count from Performance (Cluster Count Performance)) for determining the optimal k parameter for k-means clustering by using the "Optimize Parameters" operator. Unfortunately the "Performance (Combine Performance)" operator can only combine performances from a single PerformanceVector as input.

Is there another possibility to combine these two measures into a single one?

awchisholm

Hello zebra

One way is to use the Log operator to write the values to the log. You can then convert that to an example set and do what you like.

There are some examples here...

http://rapidminernotes.blogspot.co.uk/search/label/ClusterValidity

Have fun

regards

Andrew

zebra

Thanks for your answer. It's a pity that there is no easier solution using both performance vectors.