Creating Nominal to Binomial codes for a label attribute
cvsrc
New Altair Community Member
Hi all,
I am a very novice user to RapidMiner. I am trying to create a logistic regression model for a dataset where I have a few variables coded as 0/1. I wish to set one of these variables as the "label", connect it to the Nominal2Binomial Operator to transform the 0/1 nominal coded variable to a binomial variable. I wish wish to use the recoded binomial variable (variable=1 is true) in a logistic regression model.
My problems are:
1) If I set the role of the variable as "label", I dont see this variable among the attribute list for the Nominal to Binomial Operator. Is this not allowed?
2) If I set dont set the role as "label", and I perform the nominal to binomial transformation, I can see that the transformation is carried out (the metadata and Resultset show two variables created for level =0 and level = 1). However, when i connect the logistic regression (or try to connect to set role operator), I dont see the transformed variables in the attribute list.
Can anyone help me in understanding what it is that I am missing?
Let me know if posting the xml would be helpful.
Ravi
I am a very novice user to RapidMiner. I am trying to create a logistic regression model for a dataset where I have a few variables coded as 0/1. I wish to set one of these variables as the "label", connect it to the Nominal2Binomial Operator to transform the 0/1 nominal coded variable to a binomial variable. I wish wish to use the recoded binomial variable (variable=1 is true) in a logistic regression model.
My problems are:
1) If I set the role of the variable as "label", I dont see this variable among the attribute list for the Nominal to Binomial Operator. Is this not allowed?
2) If I set dont set the role as "label", and I perform the nominal to binomial transformation, I can see that the transformation is carried out (the metadata and Resultset show two variables created for level =0 and level = 1). However, when i connect the logistic regression (or try to connect to set role operator), I dont see the transformed variables in the attribute list.
Can anyone help me in understanding what it is that I am missing?
Let me know if posting the xml would be helpful.
Ravi
Tagged:
0
Answers
-
Hi Ravi,
posting the XML is (almost) always useful.
If you don't see the attribute in the Set Role operator or any other operator, you can type it in by hand. Sometimes the drop down list is not completely accurate.
For your setup, you should not use Set Role prior to Nominal to Binominal, but afterwards, if I understand correctly what you want to do.
Best,
Marius0 -
I don't see that this problem is solved by the answer. I am having the same issue in the latest version of RapidMiner.0
-
Can you please post your process xml?
Best regards,
Marius0 -
The XML is somewhat proprietary but I see that the type that comes out of "Generate Attributes" is set to "attribute_value" when i probe it but is shown as "nominal" when I have a breakpoint to display the metadata. "attribute_value" is accepted by some routines (with errors) as nominal such as "Nominal to Binomial" (complains that there is no nominal attribute but changes it to binominal anayways!) but is not *really* changed because it complains of the type "attribute_value" in many routines downstream.0
-
Hi,
the metadata (which is wrongly set to the type attribute_value in your case) is a tool which is designed to help you design a process. It does not affect the actual execution of the process, so while you will see problems in the Problems view, your process should run just fine. Nevertheless, we have an internal ticket to correct this issue in the future.
Regards,
Marco0
