This section describes the Predictor schema used for models and model configurations.
Two types of analysis models are available in Microsoft Commerce Server 2002: Prediction models and Segment models. A Prediction model is used to recommend products and fill in missing profile data. A Segment model separates users into segments which are groups of users with similar behavior and attributes. A Segment model is used to categorize a user into one or more of these segments.
These analysis models use a specified set of cases as a basis for their predictions. A case is all that is known about a previous user, and consists of property-value pairs. A property is some distinction about the world (for example, age, income, whether a particular product was purchased or not, and so on). A case typically contains transactional data and may sometimes contain demographic information.
A model configuration contains information used by the Predictor resource to build the models. This information includes pointers to the tables and views that contain the case information as well as the model build attributes. The model configuration is stored as a collection of tables in a database. There is no application programming interface (API) available to create a new model configuration. The tables must be modified directly or through the Model Configuration Wizard.
This topic includes descriptions of the required tables and entries, an example of creating a model configuration, and results from building both types of models.
This section contains:
- PredictorModelCfgs Table. Describes the available model configurations.
- PredictorDataTables Table. Serves as an index to the source data tables and an attributes table.
- Attributes Table. Describes attributes about the properties used to build the models.
- PredictorModels Table. Stores the actual models and the parameters that were used to build them.
- Predictor Schema Example. Shows an example of creating a model configuration along with the results from building both Prediction and Segment models.
See AlsoCopyright © 2005 Microsoft Corporation.
All rights reserved.