Stability of feature selection algorithms

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

Stability of feature selection algorithms

Hi to all members

I read somewhere that feature selection algorithms are considered as stable if changes in the train sample leads to the same subset of features. In Weka explorer, how can we find if the same subset of features are selected by a feature selection algorithm (i.e. Best-First Search) when train data sample is changed? My question in simple words :

 How we can make changes in training data to observe if the subset features are same or different ?

Warm regards

Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit
List etiquette: