Gini´s coefficient

[angel b.]

Hello

I am new to HL.

I am trying to use the Gini coefficient in Genetic Programming - Symbolic Regression. 

I have three doubts: 

• Each solution has a Gini coefficient (G0). When I try to simplify that solution, the coefficient changes (even when the solution has not been modified, since the simplification has been canceled. Simply by double clicking on a terminal, the coefficient changes, generally decreases). 
• After having simplified a solution and having obtained its Gini coefficient (G1 <> G0), I export the data to excel. In excel I calculate AUC and from this measure I deduce the Gini coefficient. Again the value obtained G2 is different from G1 and G0. 
• On the other hand, within the excel file, another measure of the Gini coefficient appears - average (estimated values) - that does not coincide with the previous ones.

Could you help me with the calculation method in each case? 

Thanks in advance

[Gabriel Kronberger]

Hello Angel,

thanks for bringing this to our attention. Probably there are differences in the way the value is calculated.

If you could send me an .hl file that showcases the problem and the value that you would expect then I can have a closer look at it.

Best, Gabriel

--

You received this message because you are subscribed to the Google Groups "HeuristicLab" group.

To unsubscribe from this group and stop receiving emails from it, send an email to heuristiclab...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/e022206f-b0c6-41ff-a076-fa76c15f395dn%40googlegroups.com.

[angel b.]

Hello Gabriel

I cannot post the reply message with the file. Can I send it to you in some other way?

Best Angel

[Gabriel Kronberger]

Hello Angel,

please send direct questions to sup...@heuristiclab.com

Best, Gabriel

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/56b27aae-7474-409b-a84e-5a206b161b9bn%40googlegroups.com.

[Gabriel Kronberger]

Hello Angel,

thanks for providing a file to test the behaviour. We have fixed the bug in the daily build version of HL (see https://dev.heuristiclab.com/trac.fcgi/ticket/3120)

The build with the fixed functionality should be available tomorrow.

Background: we have two different ways of calculating the Gini coefficient. One way is to use the class values produced by a classifier and the other way uses the outputs of the underlying discriminant function.

In your case the values were the result when using the class values. However, when manipulating the model in the simplified these values were overwritten with the result when using the discriminant function values.

We now report both results. With the updated HL version you should be able to open your file and when you view the results for the solution you will find the two different Gini coefficients.

Have a nice weekend,

Gabriel

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/ffdd4b5b-4a41-4b11-9704-99a8e7dfc5ab%40www.fastmail.com.

[angel beade torreiro]

Hello Gabriel,

When calculating the Gini coefficient using the results of the underlying discriminant function.

Are the above results based on the bounded values or the unbounded values?

 

Will it be possible to visualise the value of the Gini coefficient using the results of the underlying discriminant function when conducting an experiment and analysing the runs in a table?

 

Best, Angel

 

Enviado desde Correo para Windows 10

 

De: Gabriel Kronberger
Enviado: viernes, 23 de abril de 2021 18:07
Para: noreply-spamdigest via HeuristicLab
Asunto: Re: Gini´s coefficient

 

Hello Angel,

thanks for providing a file to test the behaviour. We have fixed the bug in the daily build version of HL (see https://dev.heuristiclab.com/trac.fcgi/ticket/3120)

The build with the fixed functionality should be available tomorrow.

 

Background: we have two different ways of calculating the Gini coefficient. One way is to use the class values produced by a classifier and the other way uses the outputs of the underlying discriminant function

 

In your case the values were the result when using the class values. However, when manipulating the model in the simplified these values were overwritten with the result when using the discriminant function values.

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/18c8917e-e052-4682-b59c-0731e92775f0%40www.fastmail.com.

 

Libre de virus. www.avast.com

[angel beade torreiro]

Hello Gabriel,

Sorry for the inconvenience

Which numerical integration rule is used in the calculation of the gini coefficient using the results of the underlying discriminant function?

Thanks

 

Angel

 

Enviado desde Correo para Windows 10

 

To view this discussion on the web visit https://groups.googlecom/d/msgid/heuristiclab/18c8917e-e052-4682-b59c-0731e92775f0%40www.fastmail.com.

[Gabriel Kronberger]

Hello Angel,

the code is exactly the same for calculating the Gini index for the class values and the Gini index for the outputs of the discriminant function. 

The code is here: https://dev.heuristiclab.com/trac.fcgi/browser/trunk/HeuristicLab.Problems.DataAnalysis/3.4/OnlineCalculators/NormalizedGiniCalculator.cs

The Gini coefficient is calculated for the bounded values for classification solutions produced via genetic programming (called symbolic classification in HL).

Yes, in fact both Gini values are included in the Run collection views (e.g. in the Table including all parameters and results for all runs).

Best, Gabriel

To view this discussion on the web visit https://groups.google.com/d/msgid/heuristiclab/5B5B9249-CEAA-4B5E-B1E4-FBEA0D9A8224%40hxcore.ol.