Weka Classification Probabilities

[michael....@googlemail.com]

Hi all,

I need to know about how certain my classifier is in making its decisions per instance.
I am using weka and SMO which both can (in principle) do this.
My question is, whether this use-case is currently supported in TC or whether someone knows a handy workarround/ has any experience on this.
I would like to have a report -similar to id2outcome- that gives me this info.
I already had a look on the WekaTestTask and WekaUtils where you could access the weka classifier (via cl.distributionForInstance(testData.instance(i))), this could do the trick...

Thanks in advance
-Michael

[Johannes Daxenberger]

Hi Michael,

I don't think there is any other way then to change the respective code pieces in the weka module, so you basically answered your own question :)

> I already had a look on the WekaTestTask and WekaUtils where you could access the weka classifier (via cl.distributionForInstance(testData.instance(i))), this could do the trick...

You can also have a look at the WekaOutcomeIDReport which creates the id2outcome files.

Best,
Johannes

-----Ursprüngliche Nachricht-----
Von: michael.wojatzki via dkpro-tc-users [mailto:dkpro-t...@googlegroups.com]
Gesendet: Freitag, 7. April 2017 10:01
An: dkpro-tc-users
Betreff: [dkpro-tc-users] Weka Classification Probabilities

--
You received this message because you are subscribed to the Google Groups "dkpro-tc-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dkpro-tc-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[michael....@googlemail.com]

Hi Johannes,

thanks. I will try to get this to work :)