Options
"attribute weights loader and applier"
Contributor II
Hi,
on Rapidminer Version 4.4 when I loaded attribute weights using the attribute weights loader and then applied those weights to my data, the weight of all attributes not mentioned in the attribute weights file but present in the data set was set to 0.
Using Rapidminer Version 4.5 the behaviour is opposite, it sets the weight to 1 if an attribute is not mentioned in the weights file. How can I get the behaviour of Version 4.4 back? I use it that way because in my process first all useless features are removed and the feature selection picks the useful attributes, but does not mark those removed by remove useless features as weight==0 probably because it doesn't know about them.
Peter
on Rapidminer Version 4.4 when I loaded attribute weights using the attribute weights loader and then applied those weights to my data, the weight of all attributes not mentioned in the attribute weights file but present in the data set was set to 0.
Using Rapidminer Version 4.5 the behaviour is opposite, it sets the weight to 1 if an attribute is not mentioned in the weights file. How can I get the behaviour of Version 4.4 back? I use it that way because in my process first all useless features are removed and the feature selection picks the useful attributes, but does not mark those removed by remove useless features as weight==0 probably because it doesn't know about them.
Peter
Tagged:
0
Answers
the behavior you specified is indeed not that sensible. It think we will change that with 5.0 or at least provide a parameter for that. But I took a look in the history of that operator and it has never been changed since 2007. Are you quite sure, that the behavior has changed?
Greetings,
Sebastian
sorry for the disturbance, the problem was between chair and screen. I used an AttributeWeightsApplier where I should have used an AttributeWeightSelection.
Peter