cancel
Showing results for 
Search instead for 
Did you mean: 

Eureqa variable with "-mi"

redenton
Blue LED

Hi, I am running Eureqa and there is something I don't understand about the solutions that I get. I have a number of input variables like "L", "Kp3", and "F10p73". The listed solutions are all in terms of these variables, except the last one often, though not always, appears as "F10p73-mi". Is that different from my variable name, "F10p73", in which case what does the "-mi" mean? The "-" there is not a minus sign. The whole of ""F10p73-mi" is bold white text, indicating normally a variable name. Thank you, Richard

Labels (1)
7 Replies
IraWatt
DC Motor

Hey @redenton

Looking at my Eureqa models the dash indicates its referencing a specific class within your feature. For instance in my model predicting shot number I have a feature called player_name and its referencing lamarcus aldridge creating a variable "player_name-lamarcus aldridge".

IraWatt_1-1637770394765.png

 

 

redenton
Blue LED

@IraWatt Thank you so much for responding to my question. In my case, the data structure is very simple. It's only numeric data. There are no non-numeric values in that variable. And to my knowledge there is no class structure in the data. If there is some kind of class structure, I don't think it's something that I created. I tried renaming the variable names and running it again, but I have the same issue with that particular variable, now instead of "FT" I have "FT-mi". The first line of the data file has the column names "w, cyear, syear, L, cmlt, smlt, F, FT, AE, Kp, KpA, Dst, Bysw, Bzsw, Vsw, Npsw,Pdsw, lgrhoeq" and in the columns there is only numeric data. This screenshot shows data at the top of the file. Below that is a screenshot of the Eureqa model information.Untitled.jpg

Untitled2.jpg

IraWatt
DC Motor

Has DR identified FT as a Numeric type in the data tab? does 'mi' show in the data histogram? At a guess it could be 'missing values'. I sadly haven't found much documentation on the variable names, I'm sure a Eureqa expert in the community will know.   

jas0n
Data Scientist
Data Scientist
The suffix "-mi" means that the feature is a missing value indicator. It is mentioned here: https://app.datarobot.com/model-docs/tasks/PNI2-Missing-Values-Imputed.html
 
However, I just noticed it looks like you're on a free trial so you won't have access to the modeling docs above.  The key quote from it is:
 
"In the imputation output, median value imputation is run on feature2 and feature3. The feature2-mi is the indicator column for the imputation on feature2. The feature3-mi is the indicator column for the imputation on feature3."
 
I'm also happy to discuss this further here or by e-mail if you have any follow-up questions.
 
Thanks,
Jason
redenton
Blue LED

Dear jas0n,

Thank you very much for replying about my problem. I'm not sure I understand the text that you quoted. Does it mean that feature-mi means the median value of all the values of feature in the database? In that case, I wonder why Eureqa wouldn't just use a number. How can I email you? Or it sounds like you have access to my account information. Do you have my email address? I can put it in the post if necessary.

Thank you, Richard

0 Kudos
redenton
Blue LED

Thank you for replying. No, neither of those is a variable name.

0 Kudos
jas0n
Data Scientist
Data Scientist

Hi Richard,

A missing value indicator is a bit different. It's a 0 or 1 valued feature that equals 1 if the original feature was missing in that row.

 

So if I have an original feature called X like this:

 

X

1

2

3

NA

2

4

 

Then I do missing value imputation the result will be 

X     X-mi

1     0

2     0

3     0

2.4     1

2     0

4     0 



I don't have your email but I will get it from the Community manager and email you. Mine is jason.miller@datarobot.com

Thanks,
Jason

 

0 Kudos