Scoring new datasets and Test data

[pvareschi]

In "Applied Analytics Using SAS Enterprise Miner", "Lesson 8: Model Implementation Using SAS Enterprise Miner", when scoring a new dataset ,is it correct to say it is not necessary to set up the metadata to perfectly match what done for the Training dataset? (i.e. it is sufficient to use the Basic Advisor setting)

Does the same principle apply when using an external TEST table (as done in the case study)?

[TWoodfield · Accepted solution]

The metadata for the training data tells SAS Enterprise Miner how it is allowed to use the data. The roles and levels of variables in validation, test, and score data sets are never used.

So, when creating the metadata table for individual validation, test, and score data sets, just use the basic advisor and click next until you get to the step where you specify the role of the table.