Re: Doubt about SAMOA Datasets

[Nicolas Kourtellis <nk...@gmail.com>]

Hi Eduardo,

As far as I understand, you can define your policy of what to do with the
instances arriving.
In the PrequentialEvaluator for example, all instances are used for testing
and then training.
However, the results of the evaluation are outputed every -f instances
(parameter defined at command line).

Regarding the different datasets, i assume you can join two different
datasets (for train and then test), as long as they have same attributes
structure. Then you could define the model to output test performance after
the test instances are parsed.

Hope this helps,

Nicolas



On Mon, Aug 1, 2016 at 12:53 AM, Eduardo Costa <ed...@gmail.com>
wrote:

> Dear,
> I can use two datasets: one for training and one for test with Samoa as I
> do in Weka?
> Another doubt: as SAMOA know what are the instances for testing and
> training?
>
> Thankfully,
> Eduardo.
>



-- 
Nicolas Kourtellis

Re: Doubt about SAMOA Datasets

[Eduardo Costa <ed...@gmail.com>]

Hi Nocolas!

Thanks for the explanation. But now I have other doubts :-)

1) How can I set the policy to do with the intancias arriving?
2) I read, but I can not enteder as xxx works. How can I test the instances
rating without having a classification model generated through the training
instances? Could you explain how this works? Please!
3) If I merge the two datasets (training and test) how to get just the
performance metrics for the test dataset?

Thankfully,
Eduardo.

2016-08-01 12:22 GMT-03:00 Nicolas Kourtellis <nk...@gmail.com>:

> Hi Eduardo,
>
> As far as I understand, you can define your policy of what to do with the
> instances arriving.
> In the PrequentialEvaluator for example, all instances are used for testing
> and then training.
> However, the results of the evaluation are outputed every -f instances
> (parameter defined at command line).
>
> Regarding the different datasets, i assume you can join two different
> datasets (for train and then test), as long as they have same attributes
> structure. Then you could define the model to output test performance after
> the test instances are parsed.
>
> Hope this helps,
>
> Nicolas
>
>
>
> On Mon, Aug 1, 2016 at 12:53 AM, Eduardo Costa <ed...@gmail.com>
> wrote:
>
> > Dear,
> > I can use two datasets: one for training and one for test with Samoa as I
> > do in Weka?
> > Another doubt: as SAMOA know what are the instances for testing and
> > training?
> >
> > Thankfully,
> > Eduardo.
> >
>
>
>
> --
> Nicolas Kourtellis
>