Mining Model Prediction

The Mining Model Prediction functions allow you to generate predictions from a data mining model that is stored in an Analysis Services cube. These functions can be used in T‑SQL queries, stored procedures, and client‑side applications to retrieve predictive results such as classifications, regressions, and clustering assignments.

Syntax

SELECT
    PREDICTION(
        MODEL = <MiningModelName>,
        PREDICTIVE_COL = <PredictiveColumnName>,
        INPUT = ( <column_name_1=> value1,
                  <column_name_2=> value2,
                  ... )
    ) AS PredictedValue
FROM <YourTableOrView>;

Example

This example demonstrates how to predict the CustomerSegment for a new customer using the CustomerSegmentationModel that was previously trained.

SELECT
    PREDICTION(
        MODEL = N'CustomerSegmentationModel',
        PREDICTIVE_COL = N'CustomerSegment',
        INPUT = ( Age = 35,
                  Income = 72000,
                  Gender = N'M',
                  PurchasesLastYear = 12 )
    ) AS PredictedSegment;

Remarks

• The PREDICTION function can be used only with mining models that have been deployed to an Analysis Services instance.
• All input columns must match the data type and naming used when the model was trained.
• When a column is omitted from the INPUT list, the model treats it as a missing value and applies the model's default handling (e.g., mean substitution).
• For models that support probability outputs (e.g., classification), you can retrieve probabilities using the PREDICTION_PROBABILITY function.

See also