When you use Data Mining Extensions (DMX) to define a new data mining model in Microsoft SQL Server Analysis Services, you must specify how the data mining algorithm that builds the model will use each column. You can specify a column as one of the following types:
Columns that are left unspecified in DMX are treated as input columns.
To process a model correctly, the algorithm must know which column is the key column that uniquely identifies each row, which column is the target column for creating predictions if you are creating a predictable model, and which columns to use as input columns to create the relationships that predict the target column.
Columns that are specified as the Predict type are used as both input and output columns. Columns that are specified as PredictOnly are only used as output columns. Specific algorithms may treat Predict columns differently.
For more information about the column usage types that Analysis Services supports, see Mining Model Columns.