I am currently using weka-3-2 implementing a classification with continuous class. And M5Prime regression tree and model tree are chosen to be the schemes.
Acouple of questions are raised after several running:
1. M5 will only have binary split on all attributes. How does this binary split on multiple-level nominal attribute? Does it make sense on real case? I mean if Marrital Status has Married, Single and Seperated. Why is it nessary to group Marrid and Single together after binary split?
2. What is the threshold for Infomation gain? and SDR(standard deviation reduction)?