"Obtaining gain metrics from a decision tree model"

winds

Hi --

Once I've run a DecisionTree learner, is there a way to obtain the attribute selection metric that was computed at each node (e.g. information gain or gain ratio)? 

I'm using ModelWriter to output the results, but I don't see that information there.

Also, is there an option to obtain other information such as fraction of data that was split at each node?

Any help much appreciated.

Thanks!

steffen

Hello winds

I am sure (at least with 99%) that is is not possible without digging into the code and alteration of the decision-tree-builder. Beside: It makes no sense that the requested information is stored in the model itself since it is not needed for application.

greetings,

steffen

land

Hi Wind,
Steffen is right. This isn't saved in the model at all. Of course one could change it to be saved.

Greetings,
  Sebastian