Applying predicted label value to new model application

Hello!

I hope this message finds you all well today.  I have created three models in RapidMiner 5.3, and I would like to be able to cascade them, if possible, but I have been so unsuccessful so far.  I have four labels that I ultimately want to predict, and three steps through which to reach the predictions.  I start by segmenting label 1 from 234, then 2 from 34, then 3 vs. 4.  I had this all in one model, but I found this three step process to be more accurate.

Is there a way to use the predicted value of these models to drive the next model (so that in step 2, my "1"s have been taken out of the equation and set aside)?

Thank you all for your help.

Rory