How to apply discovered formula to a new dataset?

[Claudio]

Hi everyone!
I am new to HeuristicLab. WOW what a wonderful piece of software!!

I have a question:
I can’t figure out how to use the solution founded by HeuristicLab
with a new dataset.

I’m testing genetic algorithm in a Symbolic Regression Problem (single
objective).
I’m importing my dataset from CSV, set all my inputs and output...and
then run it.

Let’s say that I’m satisfied with the discovered formula how can I
apply such a formula to a new completely dataset?

Thanks
Claudio

[Michael Kommenda]

Hello Claudio!

Thank you for the positive feedback about HeuristicLab. The
functionality you are looking for is rather new in HeuristicLab and
therefore not available in the latest release version 3.3.5. We are
planning to release 3.3.6 within the next weeks, but you can also use
the daily builds of HL [0] to apply discovered formulas on new data [0].

First of all you can't change the regression solution (a regression
solution is a mathematical formula together with the problem data) that
is the result of an algorithm run directly, because that would falsify
the algorithm results and so you have to create a copy of the solution.
The easiest way to do this, is to click on the simplify button of your
solution. Note that you can also save the resulting solution on its own.

The next step is to import your new data into a new regression problem
and to configure the problem data accordingly (especially the target
variable and the training and test partition). From this newly created
problem you have to drag and drop the problem data onto your regression
solution. This works by selecting the problem data in the new problem
and taking the innermost view host (windows label) and dragging it onto
the results list view of your regression solution. Keep in mind that all
input variables used in the discovered formula must be present in your
new dataset, because otherwise the evaluation of the model would not work.

I must admit that this workflow is currently a little bit complicated,
but this will be improved in future versions of HeuristicLab.

Best regards,
Michael

[0] http://dev.heuristiclab.com/trac/hl/core/wiki/Download

On 19/12/2011 15:37, Claudio wrote:
> Hi everyone!
> I am new to HeuristicLab. WOW what a wonderful piece of software!!
>
> I have a question:

> I can�t figure out how to use the solution founded by HeuristicLab
> with a new dataset.
>
> I�m testing genetic algorithm in a Symbolic Regression Problem (single
> objective).
> I�m importing my dataset from CSV, set all my inputs and output...and
> then run it.
>
> Let�s say that I�m satisfied with the discovered formula how can I

[Claudio]

Thanks a lot Michael!

I think I will wait few weeks for the new release 3.3.6 ;-)

Best regards
Claudio