Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Is there a faster way to do variable transformations?

Reply
Contributor
Posts: 38

Is there a faster way to do variable transformations?

Hi all,

Is there an option one can use that will enable SAS Enterprise Miners do all variable transformations and choose the best transformation for each variables? I know about the BEST transformation, but that's not suitable for every type of variable.

 

Does one need to transform one variable at a time, check distribution (or distance statistics)  to see if transformation produced desired results before moving forward with the other data preparation tasks? Does SAS EM not have a faster option? (SOrry for the repitition)

Thanks,

 

 

 

 

SAS Super FREQ
Posts: 306

Re: Is there a faster way to do variable transformations?

I"m not sure why you say Best isn't suitable for every variable - if you choose Best from the Transform Variables node, it will try several transformations for each variable (Log, Log10, Sqrt, inverse, square, exp, standardized, range) and choose the one for each variable that is most correlated with the target.  Alternatively, you can pick Maximum Normal, Maximum Correlation, Equalize, Optimal Max. Equalize for it to try several Box-Cox power transformations and then pick the best for each variable in terms of what you pick to maximize.

Ask a Question
Discussion stats
  • 1 reply
  • 116 views
  • 0 likes
  • 2 in conversation