classifier decision tree with reject option

KNIME Analytics Platform
1

Hi.
I'm working on a classifier decision tree with reject option. Practically, using java snippet, I put in a class "UNKNOWN" all traffic flows that have a decision index less than a certain threshold.

For example, I defined an index called Local Accuracy = (k_j / K)
k_j are the flows that belong to different classes and K = k_1 + k_2 +...+ k_c; with c = number of classes.

then we defined the Laplace Estimator = (k_j +1) / K + c .

Can you show me some other useful index for the rejection? perhaps easy to deploy in the predictor.

thanks,
Antonio

2

ehm.... any help ?

3

Hi, I am not sure what you mean exactly. Could you give us some more information, for example other accuracy measures you have in mind...? Cheers, Thomas

4

gabriel wrote:
Hi, I am not sure what you mean exactly. Could you give us some more information, for example other accuracy measures you have in mind...? Cheers, Thomas

I need some estimators to build a decision tree with reject option

5

In your previous post, you wrote:

Quote:
Can you show me some other useful index for the rejection?
What do you mean by "index for the rejection". I am still puzzled, is there any public reference you could point us to...

6

gabriel wrote:
In your previous post, you wrote:
Quote:
Can you show me some other useful index for the rejection?
What do you mean by "index for the rejection". I am still puzzled, is there any public reference you could point us to...

reject means put in a specified class all the patterns belonging to classes not defined inthe training set