DUMMIES/BOOLEAN VARIABLES FROM COLLECTION/LIST COLUMN

KNIME Analytics Platform
#1

I try to create dummies variable from a column containing list of values (String with comma seaprator or Collection)

A

B

A,B

null

A,B,C

A,D

A,B,D

I've 200 possible labels

I try to get boolean columns X_A, X_B, X_C X_D, etc .. containing 1 if in the row, the cell containing the specific term (i.e A, B, C, D etc...)

I cannot obtain what i want with the One To Many node, do you have any idea ?

 

 

#2

Hi Fabrice JOURDAN,

 

In order to be able to create dummy variables with a One To Many node you have to have one value per data cell, since the node doesn't work with lists or collections. 

If you have a KNIME collection (List (Collection of: String) type) you can either use a Split Collection Column or a Ungroup node to split the collections. If you have Strings with commas, you can use a Cell Splitter node with comma (",") delimiter to split the values. 

 

Best,

Anna

#3

Cell Splitter does not do "One Hot Encoding" which is necessary. Instead it respects the ordinal order of the tokens encoding "A,B" differently than "B,A". What is the solution?

#4

Cell Splitter node with comma (",")  and Ungroup seems Ok