Numeric ID not recognized as IDs by auto model

kypexin

Hi,

 

I have noticed a few times already that numeric ID is not recognized as such by auto model. 

For example, I have this type of ID in my data, obviously all values here are different (as it is indeed in ID):

 

but auto model does not detect it for some reason:

 

 

Is there any explanation for that? Does it calculate ID-ness for nominal values only?  

sgenzer

hahahaha ok I am DEFINITELY letting @IngoRM answer this one.

 

(Ingo and I have had long conversations about "ID" attributes in Auto Model...)

 

Scott

 

IngoRM

Yes, in the first version ID-ness was only calculated for nominal values.  The reason was that for numerical values, it is very likely that all values are different from each other without being an idea.  This is especially true for real values.  The next version of Auto Model will change this a bit:

 

• ID-ness for nominal attributes will be calculated as before.  Nominal attributes with an ID-ness higher than 70% are flagged with a red status.
• ID-ness for integer attributes will now also be calculated.  However, integer attributes will only be flagged with a red status if the ID-ness is higher than 99%.  This will at least cover true integer IDs without creating too many false positives where it just happens that many of the integer numbers are different...
• ID-ness for real attributes will not be calculated at all.  Their ID-ness now shows a question mark to make this clearer.

Hope this helps,

Ingo

kypexin

All clear, thanks @IngoRM