anyway you can release parts of the data or a representative subset or the
like ....? Interesting problem
On Thu, Mar 7, 2013 at 10:46 PM, Abhiram Koneru <[email protected]>wrote:
> Hello,
>
> I am trying to train a text classifier that can classify words and attach
> labels to each of them. I have a MULTI-CLASS classifier (Linear SVM). The
> classifier works well on a small training data. The problem arrives when I
> use my actual training data to run the classifier. one of my labelled set
> (lets say A) has a huge number of samples in it (~500) and the the other
> labelled sets (say B, C, D, E ~5 to 30) have very less samples when
> compared. Now this is wher it gets weird. Even if i enter an exact match
> from set A, it is labelled with B/C/D/E. I have tried changing the weights
> to 'auto' but no effect. Should I be looking at other algorithms? I have no
> clue on how to proceed.
>
> Thank you for your suggestions!
>
>
> Thanks and Regards
> --
> Abhiram Koneru
> Graduate Research Assistant
> Clemson University
> 136 Fluor Daniel Building
> Clemson, SC 29631
> Email: [email protected] Ph no: (864)643-9672
>
>
>
>
> ------------------------------------------------------------------------------
> Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester
> Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the
> endpoint security space. For insight on selecting the right partner to
> tackle endpoint security challenges, access the full report.
> http://p.sf.net/sfu/symantec-dev2dev
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester
Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the
endpoint security space. For insight on selecting the right partner to
tackle endpoint security challenges, access the full report.
http://p.sf.net/sfu/symantec-dev2dev
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
