"Using clustering result as input for K-nn / Decision tree"

joervad07

http://imgur.com/sksDyrH

This is what Ive got. It doesnt really work though.

What I want to do is use the result of the clustering for my decision tree. But right now it is classifying some random value. Not even sure how I specifiy that value that I want to predict in my model.

Thanks in advance

MartinLiebig

Hi,

first of all, i would recommend to use a (cross) validation to estimate your performance.
For using the cluster id, you can simply set the role of the newly created cluster to regular. Than it is used. You might want to build a model per cluster. A loop values helps there.

Cheers,
Martin