Module datarush.analytics

Package com.pervasive.datarush.analytics.svm.predictor

Class SVMPredictor

java.lang.Object

com.pervasive.datarush.operators.AbstractLogicalOperator

com.pervasive.datarush.operators.StreamingOperator

com.pervasive.datarush.operators.ExecutableOperator

com.pervasive.datarush.analytics.util.AbstractPredictor

com.pervasive.datarush.analytics.svm.predictor.SVMPredictor

All Implemented Interfaces:

LogicalOperator

public final class SVMPredictor
extends AbstractPredictor

Operator responsible for classification based on a SVM PMML model. This supports either CSVC SVMs or one-class SVMs. We distinguish the two cases by the presence of PMMLModelSpec.getTargetCols(). If there are zero target columns, it is assumed to be a one-class SVM. Otherwise, there must be exactly of column of type TokenTypeConstant.STRING, in which case it is a CSVC SVM.

For CSVC SVMs, the PMML is expected to contain SupportVectorMachine's with SupportVectorMachine#getTargetCategory() and SupportVectorMachine#getAlternateTargetCategory() populated. Each of the SVM's are evaluated, adding a "vote" to either target category or alternate target category. The predicted value is that receiving the most votes.

For one-class SVMs, target category and alternate target category will be ignored. The result will either be "-1" if the SupportVectorMachine evaluated to a number less that zero or "1" if greater than zero.

NOTE: this operator is non-parallel

Field Summary

Fields

Modifier and Type	Field	Description
static final String	FIELD_WINNER	Constant for the winner field.

Constructor Summary

Constructors

Constructor	Description
SVMPredictor()	Creates a new SVMPredictor.

Method Summary

Modifier and Type	Method	Description
protected void	computeMetadata(StreamingMetadataContext ctx)	Default implementation of computeMetadata.
protected void	execute(PMMLModel pmml, RecordValued input, ScalarSettable[] predictedFields)	Called to perform prediction.
protected RecordTokenType	predictedType(PMMLModelSpec modelSpec)	Given the model spec, returns the predicted type.

Methods inherited from class com.pervasive.datarush.analytics.util.AbstractPredictor

execute, getInput, getModel, getOutput, pushPrediction, stepNext

Methods inherited from class com.pervasive.datarush.operators.ExecutableOperator

cloneForExecution, getNumInputCopies, getPortSettings, handleInactiveOutput

Methods inherited from class com.pervasive.datarush.operators.AbstractLogicalOperator

disableParallelism, getInputPorts, getOutputPorts, newInput, newInput, newOutput, newRecordInput, newRecordInput, newRecordOutput, notifyError

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

FIELD_WINNER

public static final String FIELD_WINNER

Constant for the winner field.

See Also:

• Constant Field Values

Constructor Details

SVMPredictor

public SVMPredictor()

Creates a new SVMPredictor.

Method Details

computeMetadata

protected void computeMetadata(StreamingMetadataContext ctx)

Description copied from class: AbstractPredictor

Default implementation of computeMetadata.

1. Output type is set to input type plus predictedType
2. Input data ordering ( if ordered ) is preserved
3. Input data partitioning ( if partitioned ) is preserved

Overrides:

computeMetadata in class AbstractPredictor

Parameters:

ctx - the context

predictedType

protected RecordTokenType predictedType(PMMLModelSpec modelSpec)

Description copied from class: AbstractPredictor

Given the model spec, returns the predicted type. This should not include the input type ( the input is automatically prepended to the type that is returned )

Specified by:

predictedType in class AbstractPredictor

Parameters:

modelSpec - the model metadata

Returns:

the predicted type

execute

protected void execute(PMMLModel pmml,
 RecordValued input,
 ScalarSettable[] predictedFields)

Description copied from class: AbstractPredictor

Called to perform prediction. Subclasses are expected to loop over the input by calling AbstractPredictor.stepNext(). For each row of input, subclasses should first set the predicted values in the predictedFields array and then invoke AbstractPredictor.pushPrediction(). Subclasses should not invoke pushEndOfData since that is automatically handled by the base class.

Specified by:

execute in class AbstractPredictor

Parameters:

pmml - The input PMML model

input - The input data

predictedFields - An array of fields that reference the predicted field locations. The array positionally corresponds to the type returned by AbstractPredictor.predictedType(PMMLModelSpec).