Before and After updating the Rapidminer studios into 9.6 have different prediction accuracy results. Which one is correct, and what made them different?
Have you strictly set the same option(s) in both cases ? Have you enable the feature selection option ? if yes can you check that the same feature set is identical in both cases ?
Edit :
An other hypothesis : RapidMiner 9.6 introduce the notion of gain. (You have to set a cost matrix in the menu of AutoModel). In RapidMiner 9.6 the model are built to optimize the gain(s) and not the accuracy directly. So the models are built differently. Thanks to the RM staff to correct me, if I'm wrong...
Answers
Have you strictly set the same option(s) in both cases ?
Have you enable the feature selection option ? if yes can you check that the same feature set is identical in both cases ?
Edit :
An other hypothesis : RapidMiner 9.6 introduce the notion of gain. (You have to set a cost matrix in the menu of AutoModel).
In RapidMiner 9.6 the model are built to optimize the gain(s) and not the accuracy directly. So the models are built differently.
Thanks to the RM staff to correct me, if I'm wrong...
Regards,
Lionel