"Decision Tree - Multiple Target Column"

j_pries

Hey,

I´m a german student and on a study research for automated selection of grippers according to input-properties of objects. I want to use a decision tree.

The problem is that there are more than one possible gripping principles for an object. For example, a cube can be gripped by clamping and suction grippers. Is it possible to use more than one target column in training data or how can solve the problem?

Thanks for Your help!

Telcontar120

You can't have more than one target column (called a label in RapidMiner) at the same time in RapidMiner.

However, you could build multiple DT models using different labels and then combine their predictions. Or you could create a new single label that incorporates information from mulitple original separate target columns using Generate Attributes (e.g., maybe you create a single label to indicate if an object can be picked up using neither method, one method, or both methods).

JEdward

Here's an example of using Loop Labels.

< ?xml version = " 1.0 " encoding = " utf - 8 " ?> <过程版本sion="9.0.002">

































< portSpacing端口= "苏rce_in 1" spacing="0"/>










Set Role of label columns to label1, label2









< portSpacing端口= "苏rce_example set" spacing="0"/>



Don;t forget to use validation


















< portSpacing端口= "苏rce_single" spacing="0"/>




You can combine models or score them separately like this example.






< portSpacing端口= "苏rce_input 1" spacing="0"/>