A probabilistic model contains reference data values and label values. The reference data values represent the data on an input port that you connect to the transformation. The label values describe the types of information that the reference data values contain. You assign a label to each reference data value in the model.

To link the reference data values to the labels in a probabilistic model, you compile the model. The compilation process generates a series of logical associations between the data values and the labels. When you run a mapping that reads the model, the Data Integration Service applies the model logic to the transformation input data. The Data Integration Service returns the label that most accurately describes the input data values.

You create a probabilistic model in the Developer tool. The Model repository stores the probabilistic model object. The Developer tool writes the data values, the labels, and the compilation data to a file in the Informatica directory structure.