decision fusion

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

decision fusion

shaimaa
I have two sets of features concerning a specific data and I want to make a decision fusion. Can weka help.
thanks.

_______________________________________________
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: decision fusion

Eibe Frank-2
Administrator
Yes. The easiest way to achieve this is to join the two sets of features into a single table (e.g., in an ARFF file). Then use Stacking with two FilteredClassifier objects as the base classifiers: the first FilteredClassifier should use the Remove filter to remove all attributes that are not in your first set of features; the second FilteredClassifier should remove the other features (leaving the class attribute in both sets, obviously).

ClassificationViaRegression with M5P as the base learner has been found (empirically) to be a good *meta* learner in Stacking. You will need to experiment with different base learners in the two FilteredClassifier objects but RandomForest is generally a classifier that gives reasonable results.

Of course, you should compare to building a RandomForest (or whichever base classifier you end up using) on the union of all features to see whether there is any benefit in your multi-view approach based on two separate feature sets.

Cheers,
Eibe

> On 30/07/2019, at 10:35 PM, shaimaa hagras <[hidden email]> wrote:
>
> I have two sets of features concerning a specific data and I want to make a decision fusion. Can weka help.
> thanks.
> _______________________________________________
> 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