How I can avaliate different classifier's model of a same function Classifier

[Lyiu]

I would like testing a several multilayer Perceptron's model saved by me, and compare and avaliate the outputs. How I can do this in a Knowledge flow? It's possible?

[Peter Reutemann-3]

> I would like testing a several multilayer Perceptron's model saved by me, and
> compare and avaliate the outputs. How I can do this in a Knowledge flow?
> It's possible?

I don't think so that you can do that in the KnowledgeFlow. You can
use the Experimenter for statistical evaluation (and comparison) of
classifiers.

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Lyiu]

>I don't think so that you can do that in the KnowledgeFlow. You can
>use the Experimenter for statistical evaluation (and comparison) of
>classifiers.

But I needed a confusion matrix, and the Experimenter has not the confusion matrix(I think so not). Have another way to do this?

Thank you...

[Peter Reutemann-3]

>>I don't think so that you can do that in the KnowledgeFlow. You can
>>use the Experimenter for statistical evaluation (and comparison) of
>>classifiers.
>
> But I needed a confusion matrix, and the Experimenter has not the confusion
> matrix(I think so not). Have another way to do this?

Nope, don't think so. Apart from creating several flows with the
different MLP setups and then manually comparing the evaluation
output.

Why is the confusion matrix so important? Isn't the number of
incorrectly classified instances, AUC, various errors, etc. not
sufficient?

Cheers, Peter
--
Peter Reutemann, Dept. of Computer Science, University of Waikato, NZ
http://www.cs.waikato.ac.nz/~fracpete/           Ph. [phone]

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@...
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

[Lyiu]

>Nope, don't think so. Apart from creating several flows with the
>different MLP setups and then manually comparing the evaluation
>output.

>Why is the confusion matrix so important? Isn't the number of
>incorrectly classified instances, AUC, various errors, etc. not
>sufficient?

>Cheers, Peter

I needed to analyze the output of confusion matrix too...but your help was very apreciate...
My problem just was resolved...
Thank you.....