preprocessing for multiple models
meliniak
New Altair Community Member
i want to train several models. there is one training data set, but each model is provided with the examples that match certain criteria. in my case, that criterion is simply a value of an attribute. for example, first decision tree is trained with examples that 'temperature' attribute value is 'very cold', second decision tree is trained with examples that 'temperature = cold', and so on.
needless to say, each model is supposed to classify only examples that match the same criteria as they were trained with. when models are ready, i want to be able to use them in my java code.
what approach would you suggest? notice that there's some preprocessing (i.e. example selection for appropriate models), so the same preprocessing needs to be applied before the classification. is there any way to somehow embed preprocessing operators in a single model? or other operator? or do i need simply to create similar process, with the very same preprocessing operators, 'apply model' operators, and use whole Process class instead of Model one?
can you elaborate on this topic? any hints higly appreciable
needless to say, each model is supposed to classify only examples that match the same criteria as they were trained with. when models are ready, i want to be able to use them in my java code.
what approach would you suggest? notice that there's some preprocessing (i.e. example selection for appropriate models), so the same preprocessing needs to be applied before the classification. is there any way to somehow embed preprocessing operators in a single model? or other operator? or do i need simply to create similar process, with the very same preprocessing operators, 'apply model' operators, and use whole Process class instead of Model one?
can you elaborate on this topic? any hints higly appreciable
0
Answers
-
Hi,
I answered this pretty much in your other thread here: http://rapid-i.com/rapidforum/index.php/topic,3867.0.html
Regards,
Marco0 -
I just desperately needed answer. Thanks again!0
