I was looking at it from a similar point of view to outlier detection using
a single class SVM and wondering if the technique was applicable to
decision to try and "explain" the data rather that perform
classification/regression. Haven't looked at the literature but my point of
reference was decision tree clustering of HMMs representing fragments of
speech that are not tagged, i.e., building up the decision tree without
class labels using BIC/MDL and then use it for classification. Like
clustering.
On Tue, Feb 28, 2012 at 4:24 PM, Olivier Grisel <[email protected]>wrote:
> 2012/2/28 Nicholas Pilkington <[email protected]>:
> > Is there any easy was build up a decision tree for the purposes of
> > clustering instead of for classification/regression?
>
> To me decision trees is intrinsically supervised algorithm. The split
> of node of the tree is computed by a entropy measure w.r.t. the target
> variable class distribution. How would you do that in an unsupervised
> clustering case? Do you have any reference in the literature that
> motivates this question?
>
> --
> Olivier
> http://twitter.com/ogrisel - http://github.com/ogrisel
>
>
> ------------------------------------------------------------------------------
> Keep Your Developer Skills Current with LearnDevNow!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-d2d
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
--
Nicholas C.V. Pilkington
University of Cambridge
------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
