You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mahout.apache.org by Mahmood Naderan <nt...@yahoo.com> on 2014/03/31 16:20:01 UTC

Using "split" without partitioning the data to train/test

Hi,
In an old Mahout, I used wikipediaDataSetCreator on an input to create the training data
    
    mahout wikipediaDataSetCreator -i 
wiki-tr/chunks -o tr-input -c labels.txt 

and then fed the tr-input to the trainclassifier using

    mahout trainclassifier -i tr-input -o wikimodel


Now, in Mahout 0.9, I see some examples that create 80% of the input file as training model using "split"

    mahout split -i input-vectors --trainingOutput tr-vectors --testOutput ts-vectors --randomSelectionPct 20

My question is how can I use "split" to split the input without partitioning it to train and test parts? I want to use one file as training input and the other file as the test input.


 
Regards,
Mahmood

Re: Using "split" without partitioning the data to train/test

Posted by Mahmood Naderan <nt...@yahoo.com>.
Yeah you are right. I have to ignore that command

 
Regards,
Mahmood
On Monday, March 31, 2014 6:56 PM, Suneel Marthi <su...@yahoo.com> wrote:
 


Sent from my iPhone

> On Mar 31, 2014, at 4:20 PM, Mahmood Naderan <nt...@yahoo.com> wrote:
> 
> Hi,
> In an old Mahout, I used wikipediaDataSetCreator on an input to create the training data
>    
>     mahout wikipediaDataSetCreator -i 
> wiki-tr/chunks -o tr-input -c labels.txt 
> 
> and then fed the tr-input to the trainclassifier using
> 
>     mahout trainclassifier -i tr-input -o wikimodel
> 
> 
> Now, in Mahout 0.9, I see some examples that create 80% of the input file as training model using "split"
> 
>     mahout split -i input-vectors --trainingOutput tr-vectors --testOutput ts-vectors --randomSelectionPct 20
> 
> My question is how can I use "split" to split the input without partitioning it to train and test parts? I want to use one file as training input and the other file as the test input.

So why use 'split'?  Separate out the test and training files. 

> 
> 
>  
> Regards,
> Mahmood

Re: Using "split" without partitioning the data to train/test

Posted by Suneel Marthi <su...@yahoo.com>.

Sent from my iPhone

> On Mar 31, 2014, at 4:20 PM, Mahmood Naderan <nt...@yahoo.com> wrote:
> 
> Hi,
> In an old Mahout, I used wikipediaDataSetCreator on an input to create the training data
>     
>     mahout wikipediaDataSetCreator -i 
> wiki-tr/chunks -o tr-input -c labels.txt 
> 
> and then fed the tr-input to the trainclassifier using
> 
>     mahout trainclassifier -i tr-input -o wikimodel
> 
> 
> Now, in Mahout 0.9, I see some examples that create 80% of the input file as training model using "split"
> 
>     mahout split -i input-vectors --trainingOutput tr-vectors --testOutput ts-vectors --randomSelectionPct 20
> 
> My question is how can I use "split" to split the input without partitioning it to train and test parts? I want to use one file as training input and the other file as the test input.

So why use 'split'?  Separate out the test and training files. 
> 
> 
>  
> Regards,
> Mahmood