How to use prediction task after "ExperimentSaveModel" task

[lakhani...@gmail.com]

I have task of predicting my data set with a save model. I have saved my model using:

"https://github.com/dkpro/dkpro-tc/blob/master/dkpro-tc-examples/src/main/java/org/dkpro/tc/examples/multi/document/MekaSaveAndApplyModelMultilabelDemo.java"

can you point me where could I get an example for predicting a data with saved model?

the problem in multilabel demo sample is that the classification outcome is annotated, however I want a weighted outcome for the multi-class classification i.e. class 1 has 0.6, class 2 has 0.3 and so on..

Is there a way to add report like traintest report in the pipeline?

[Richard Eckart de Castilho]

> On 19.07.2016, at 18:19, lakhani...@gmail.com wrote:
>
> I have task of predicting my data set with a save model. I have saved my model using:
>
> "https://github.com/dkpro/dkpro-tc/blob/master/dkpro-tc-examples/src/main/java/org/dkpro/tc/examples/multi/document/MekaSaveAndApplyModelMultilabelDemo.java"
>
> can you point me where could I get an example for predicting a data with saved model?

Maybe this helps...

https://github.com/dkpro/dkpro-core/blob/master/dkpro-core-flextag-asl/src/main/java/de/tudarmstadt/ukp/dkpro/core/flextag/FlexTag.java

... but I hope somebody else hast a better pointer/example.

> the problem in multilabel demo sample is that the classification outcome is annotated, however I want a weighted outcome for the multi-class classification i.e. class 1 has 0.6, class 2 has 0.3 and so on..
>
> Is there a way to add report like traintest report in the pipeline?

I don't think reports can easily be used outside a DKPro TC experimental setup.

Cheers,

-- Richard

[Johannes Daxenberger]

In the case you mention, most of the “prediction magic” happens in the LoadModelConnectorWeka class: https://github.com/dkpro/dkpro-tc/blob/master/dkpro-tc-ml-weka/src/main/java/org/dkpro/tc/weka/task/serialization/LoadModelConnectorWeka.java

This is where the actual prediction takes place. You would probably have to adapt this class add weighting to your outcome votes.

Best,
Johannes

Am 19.07.16 21:27 schrieb "dkpro-t...@googlegroups.com im Auftrag von Richard Eckart de Castilho" <dkpro-t...@googlegroups.com im Auftrag von richard...@gmail.com>:

--
You received this message because you are subscribed to the Google Groups "dkpro-tc-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dkpro-tc-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[lakhani...@gmail.com]

Thank you all for the response.