The OBE exercises show you both how to build models and then to apply (Score) new data using the built models.
A model is always built using some form of input data, so it is built specifically with that form of data in mind.
It is not a generic model at all.
When you apply a model you provide data in the same format as the original data.
In the case of a Classification or Regression model, you are applying the model to generate a prediction on new data that conforms with the build data provided to the model.
The online help provides details on all the models that are available.
Data Miner uses the data mining pl/sql packages (package name DBMS_DATA_MINING) to create and test models.
There are also sql data mining prediction functions as well.
Data Miner only generates the sql for the transformation lineage defined in the workflow at the moment.
The next release will have the ability to generate the script for the entire workflow.
You can get a better appreciation for how to use the pl/sql code by viewing the sample code provided by the DB release.
Here is a pointer to a zipped file of the sample programs.