Re: Recurrent Neural Network for Time Series: Error message

The RelationalInstanceIterator is useful for time series classification or
time series regression. I don't think it is the right iterator for
forecasting. To use the RelationalInstanceIterator, each time series must
be represented as the value of a relation-valued attribute, see

https://deeplearning.cms.waikato.ac.nz/user-guide/data/#relationalinstanceiterator

The airline data does not have a relational attribute.

There may be no way to easily use an LSTM (or other recurrent model) for
time series *forecasting* in WEKA at the moment. You could give

https://github.com/braun-steven/weka-rnn-forecaster

a try, but I don't know whether it works with the latest versions of the
packages that it depends on (wekaDl4j and timeSeriesForecasting). I suspect
it doesn't.

Cheers,
Eibe

On Tue, 26 Sept 2023 at 22:33, Carlos Diaz <[email protected]> wrote:

> Hello everyone.
>
> I tried to run a model with a time series using wekaDeeplearning4j but I
> received the following error message: "Problem evaluating forecaster: The
> attribute at index <0 is not relational".
>
> My configuration was as follows:
>
> Data: airline_passengers
> Test options: cross-validation 10 folds
> Classifier: Dl4jMlpClassifier
> Class: (Num) passenger_numbers
> layer specification: LSTM - activation function: ActivationTanH, gate
> activation function: ActivationTanH, the other parameters with default
> settings
>                                  RnnOutputLayer - activation function:
> ActivationTanH, loss function: LossMSE, the other parameters with default
> settings
> instance iterator: RelationalInstanceIterator - truncation length: 144,
> the other parameters with default settings
> All the other parameters with default settings
>
> With the Forecast package, I received the same error message.
>
> I tried changing the relational attribute index from 0 to 1 but I got the
> same message. When using 2 or more I receive the message: "Problem
> evaluating classifier: Index 2 out of bounds for length 2"
>
> Any idea of how to solve it?
>
> Thank you.
>
> Carlos
>
> P.D.: I have a Mac and a Windows computer and the same happened using both
> systems.
> _______________________________________________
> Wekalist mailing list -- [email protected]
> Send posts to [email protected]
> To unsubscribe send an email to [email protected]
> To subscribe, unsubscribe, etc., visit
> https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
> List etiquette:
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
>

_______________________________________________
Wekalist mailing list -- [email protected]
Send posts to [email protected]
To unsubscribe send an email to [email protected]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html