Prerequisites for Retraining a Model
Before you can retrain
a model, complete the following prerequisites:
-
If you want
to retrain the project champion model, ensure that the champion model
is set.
For more information,
see Champion Models.
-
Verify that
the training data set that you want to use as the training data source
has been registered in the SAS Metadata Repository, and is available
in the Data category view.
-
Verify that the appropriate project
and model properties are set:
Classification Model Project Properties
-
-
-
-
Output event probability variable
Prediction Model Project Properties
-
-
-
Output prediction variable
-
-
Verify that the retrain file that
is specified in the model template exists in the list of model files.
The retrain file must appear on the Model Properties page
for the model that you want to retrain. Ensure that the content is
correct.
-
In order to retrain SAS/STAT linear
models, you must modify the training code in the train code file (for
example, batch.sas).
To edit the model train
code file:
-
-
On the
Model
Properties page, select
Advanced Model Files.
-
Double-click the train
code file (batch.sas) to open it.
-
Select the
Edit check
box and make the following modifications to the code.
-
Replace the name of the training
data set with the macro variable &_MM_InputDS
.
-
Replace the name of the outmodel
with smmmodel.outmodel
.
-
Make sure that the name of the
itemstore is work.itemstore
.
SAS/STAT Logistic Model Training Code
proc logistic data=&_MM_InputDS outmodel=smmmodel.outmodel;
class mstr_cstmr_age_grp days_claim_grp odm_last_veh_prchsd_grp
highest_edu_grp;
model bcr_ind=mstr_cstmr_age_grp days_claim_grp odm_last_veh_prchsd_grp
highest_edu_grp;
;
store work.itemstore;
quit;
-
Click
.
Copyright © SAS Institute Inc. All Rights Reserved.
Last updated: February 22, 2017