When using the Rapid Deployment to deploy PMML scripts for a tree-based algorithm such as C&RT and CHAID to a test data, you get missing value for prediction results in some cases. It is likely due to missing value in the test data. Currently, the PMML deployment requires all input variables (you specified as predictors when building the model for PMML scripts) to be non-missing, in order to generate a prediction for that test case. The ideal condition is that PMML deployment only generates missing prediction for cases with missing values among those predictors that are actually used as split variables in the deployment model.
 

WORKAROUND:

1. Check your testing data to see if any missing value in those predictors that are used in the splitting rules of the final deployment model
2. If your don't find missing value in step 1, select the option "Predict case(s) with missing data in inputs" during the deployment process

3. If you find missing value in step 1, decide if you still want to make prediction for such condition. By selecting "Predict case(s) with missing data in inputs", you will get prediction results even in this condition for tree-based models.