May be this will clarify problem -
I have unbalanced class data in which I am performing SVM classification.
Classification with equal weight for both class gives very
low sensitivity etc. (misclassification of low data class).
I tried give different class weight but I am not sure how to do
it automatically?
As I have to make several classification models it is not possible to give
class weight for every mode manually.
So I thought to give a try to oversampling or undersampling methods.
But from the answers it seems its not a best choice and make problem more
complicated.
So now I rephrase questions (may be I should make new question!!!!)--
How to chose class weight automatically in unbalanced classes data
classification?
Is it possible to do it via function GridSearchCV?
On Wed, Jul 25, 2012 at 5:33 PM, Jaques Grobler <[email protected]>wrote:
> Could you perhaps clarify what you would want to use this for?
>
> AFAIK the scikit-learn doesn't explicitly implement those techniques but
> perhaps it can be done
> another way through some of the existing implementations, like useing
> class_weights or re-calibration - if it has to do with imbalanced data for
> classification for example.
>
> Hope this helps.
> J
>
>
>
> 2012/7/25 Sheila the angel <[email protected]>
>
>> yes that is what I am looking for
>> http://en.wikipedia.org/wiki/Oversampling_and_undersampling_in_data_analysis
>> Also in the reference
>> http://nd.edu/~dial/papers/SPRINGER05.pdf<http://nd.edu/~dial/papers/SPRINGER05.pdf>
>> (though I haven't read it fully)
>>
>> On Wed, Jul 25, 2012 at 4:16 PM, Jaques Grobler
>> <[email protected]>wrote:
>>
>>> @Sheila
>>>
>>> Are you referring to this type of thing?
>>>
>>>
>>> http://en.wikipedia.org/wiki/Oversampling_and_undersampling_in_data_analysis
>>>
>>>
>>>
>>> 2012/7/25 Gael Varoquaux <[email protected]>
>>>
>>>> On Wed, Jul 25, 2012 at 02:13:41PM +0200, Sheila the angel wrote:
>>>> > I would like to know are oversampling and undersampling methods
>>>> > implemented in sklearn???
>>>>
>>>> I don't know what you mean by oversampling and undersampling. Could you
>>>> detail.
>>>>
>>>> Gael
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Live Security Virtual Conference
>>>> Exclusive live event will cover all the ways today's security and
>>>> threat landscape has changed and how IT managers can respond.
>>>> Discussions
>>>> will include endpoint security, mobile security and the latest in
>>>> malware
>>>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>>>> _______________________________________________
>>>> Scikit-learn-general mailing list
>>>> [email protected]
>>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>>
>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Live Security Virtual Conference
>>> Exclusive live event will cover all the ways today's security and
>>> threat landscape has changed and how IT managers can respond. Discussions
>>> will include endpoint security, mobile security and the latest in malware
>>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>>> _______________________________________________
>>> Scikit-learn-general mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> Live Security Virtual Conference
>> Exclusive live event will cover all the ways today's security and
>> threat landscape has changed and how IT managers can respond. Discussions
>> will include endpoint security, mobile security and the latest in malware
>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
