Measure the similarity/difference between two decision trees produced by J48

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Measure the similarity/difference between two decision trees produced by J48

Φερετζάκης Γεώργιος

Hello,

I'd like to ask if there is an option/tool that can measure or even better identify the difference between two unpruned decision trees that produced from J48?


Thank you very much.


Best Regards


George Feretzakis

PhD(c) in Machine Learning


_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Measure the similarity/difference between two decision trees produced by J48

Eibe Frank-2
Administrator
Perhaps the only way to currently compare trees in WEKA is the generic method applicable to all classifiers, based on comparing predictions: label the dataset X using classifier A that was built from X to get a new dataset X', build classifier B on X and evaluate it using X’ as the test set.

Example using the console of the operating system, comparing J48 to REPTree:

$ java weka.Run .AddClassification -W .J48 -c last -classification -remove-old-class < ~/datasets/UCI/iris.arff | java weka.Run .RenameAttribute -R last -replace "class" > iris.J48_classifications.arff
$ java weka.Run .REPTree -t ~/datasets/UCI/iris.arff -T iris.J48_classifications.arff

The kappa statistic in the output is one common measure of agreement between classifiers.

Cheers,
Eibe

> On 29/01/2019, at 12:20 AM, Φερετζάκης Γεώργιος <[hidden email]> wrote:
>
> Hello,
> I'd like to ask if there is an option/tool that can measure or even better identify the difference between two unpruned decision trees that produced from J48?
>
> Thank you very much.
>
> Best Regards
>
> George Feretzakis
> PhD(c) in Machine Learning
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
Reply | Threaded
Open this post in threaded view
|

Re: Measure the similarity/difference between two decision trees produced by J48

Φερετζάκης Γεώργιος

Eibe,

Thank you very much.


Best,

George 




From: [hidden email] <[hidden email]> on behalf of Eibe Frank <[hidden email]>
Sent: Tuesday, January 29, 2019 3:13 AM
To: Weka machine learning workbench list.
Subject: Re: [Wekalist] Measure the similarity/difference between two decision trees produced by J48
 
Perhaps the only way to currently compare trees in WEKA is the generic method applicable to all classifiers, based on comparing predictions: label the dataset X using classifier A that was built from X to get a new dataset X', build classifier B on X and evaluate it using X’ as the test set.

Example using the console of the operating system, comparing J48 to REPTree:

$ java weka.Run .AddClassification -W .J48 -c last -classification -remove-old-class < ~/datasets/UCI/iris.arff | java weka.Run .RenameAttribute -R last -replace "class" > iris.J48_classifications.arff
$ java weka.Run .REPTree -t ~/datasets/UCI/iris.arff -T iris.J48_classifications.arff

The kappa statistic in the output is one common measure of agreement between classifiers.

Cheers,
Eibe

> On 29/01/2019, at 12:20 AM, Φερετζάκης Γεώργιος <[hidden email]> wrote:
>
> Hello,
> I'd like to ask if there is an option/tool that can measure or even better identify the difference between two unpruned decision trees that produced from J48?
>
> Thank you very much.
>
> Best Regards
>
> George Feretzakis
> PhD(c) in Machine Learning
> _______________________________________________
> Wekalist mailing list
> Send posts to: [hidden email]
> To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
> List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html

_______________________________________________
Wekalist mailing list
Send posts to: [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html