PredModelBuilder Object

Ee811805.c++_off(en-US,CS.10).gifEe811805.vb_on(en-US,CS.10).gif

Use this object to control the creation of analysis models using multiple prioritized build threads.

ProgID:   Commerce.PredictorModelBuilder (Externally creatable)
COM Class Name:   PredModelBuilder
COM Interface Name:   IPredModelBuilder
Interface ID Constant:   IID_IPredModelBuilder
Header File:   PredictorService.h
Type Library Name:   Microsoft Commerce 2000 Predictor Service Type Library
EXE Name:   PredServ.exe
Threading Model:   Both

In C++, use the IPredModelBuilder interface to access the properties and methods of the PredModelBuilder object.

The properties of the PredModelBuilder object are shown in the following table.

Property Type Description
dModelInfo IDictionary Retrieves information on the specified running model.

This property is read-only.

Priority enum Stores the build thread priority.

This property is read/write.

slRunningModels ISimpleList Retrieves a list of running models.

This property is read-only.

Status enum Indicates the build thread status.

This property is read-only.

The methods of the PredModelBuilder object are shown in the following table.

Method Description
Pause Pauses a running build thread.
Resume Resumes a paused build thread.
Start Sets the thread identifier for a build and starts a new build thread.
Stop Stops a running build thread.
StopAllBuilds Stops all currently running build threads.

Remarks

Two types of analysis models are available in Microsoft Commerce Server 2000. A Prediction model is a collection of decision trees (called a dependency net) and is used either to produce a ranked list of recommendations, such as products to advertise, or to fill in missing user profile information, such as age. A Segment model is based on the process of clustering (segmentation) and is used to group similar users into segments of a population.

An analysis model is associated with a model configuration. More than one analysis model can be associated with the same model configuration. The model configuration contains attribute and case information used to build the analysis model. A case is all the information known about a user, such as profile data and transaction data. A typical model uses between 10,000 and 20,000 cases with a resulting size of between 100 kilobytes (KB) and 10 megabytes (MB).

The PredictorServiceAdmin object contains lists of the available analysis models and model configurations. The PredictorServiceSiteAdmin object provides methods for maintaining the analysis models and model configurations. The PredictorClient object loads the appropriate analysis model and makes the actual prediction.

The generation of an analysis model occurs in an individual build thread. A single instance of the PredModelBuilder object may control multiple simultaneous build threads. If a problem occurs in a build thread, that specific thread may be paused or terminated.

See Also

Prediction Objects

Predictor Schema

PredictorClient Object

PredictorServiceAdmin Object

PredictorServiceSiteAdmin Object

Running the Predictor Resource


All rights reserved.