When you run a mapping that includes a model, the Data Integration Service applies the compiled model logic to the transformation input data. The Data Integration Service does not read the data values or the labels in the model when the mapping runs.

You can optionally remove the data values and the labels from a probabilistic model or a classifier model. For example, you might decide to remove sensitive data or proprietary data from a model. You can remove individual data values and labels in the Developer tool. You can remove all data values and labels when you export a model from the Model repository.

When you remove one or more data values or labels from a model, the compiled model logic no longer represents the current data in the model file. To synchronize the model logic and the data values and labels, compile the model again. Do not compile the model if you want to maintain the current model logic.

To protect the data in a classifier model or a probabilistic model, back up the model file in the Informatica directory structure. Back up the file before you remove all the data values and labels from a model.

Find the model files in the Content Management Service host machine.

Probabilistic model files have the following default location and file name extension:

Classifier model files have the following default location and file name extension:

If you upgrade the Informatica installation, you might need to compile the probabilistic models and classifier models before you use the models in a mapping. If a model does not contain any data, replace the current file in the Informatica directory structure with the backup file that contains the data.