weka.distributed

Class WekaScoringMapTask

java.lang.Object

weka.distributed.WekaScoringMapTask

All Implemented Interfaces:

java.io.Serializable

public class WekaScoringMapTask
extends java.lang.Object
implements java.io.Serializable

Map task for scoring data using a model that has been previously learned. Handles both classifiers and clusterers. Builds a mapping between incoming fields and those that the model expects. Degrades gracefully (via missing values) when model fields are missing from the incoming data or there are type mismatches.

Version:

$Revision: 12013 $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
WekaScoringMapTask()

Method Summary

Modifier and Type	Method and Description
double[][]	finalizeBatchPrediction() Finish off the last partial batch (if any).
java.lang.String	getMissingMismatchAttributeInfo() Get a string summarizing missing and type mismatches between the incoming data and what the model expects
java.util.List<java.lang.String>	getPredictionLabels() Get a list of labels that the model can predict
boolean	isBatchPredictor() Returns true if the underlying model is a BatchPredictor
boolean	modelIsAClassifier() Returns true if the underlying model is a classifier
boolean	modelIsUsingStringAttributes() Returns true if model is using string attributes
double[]	processInstance(weka.core.Instance inst) Process (score) an instance
double[][]	processInstanceBatchPredictor(weka.core.Instance inst) Process an instance.
void	setModel(java.lang.Object model, weka.core.Instances modelHeader, weka.core.Instances dataHeader) Set the model to use
void	updateModel(java.lang.Object model) Update the underlying model for this scoring task

Methods inherited from class java.lang.Object

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

Constructor Detail

WekaScoringMapTask

public WekaScoringMapTask()

Method Detail

setModel

public void setModel(java.lang.Object model,
                     weka.core.Instances modelHeader,
                     weka.core.Instances dataHeader)
              throws DistributedWekaException

Set the model to use

Parameters:

model - the model to use

modelHeader - the header of the training data used to train the model

dataHeader - the header of the incoming data

Throws:

DistributedWekaException - if more than 50% of the attributes expected by the model are missing or have a type mismatch with the incoming data

updateModel

public void updateModel(java.lang.Object model)
                 throws DistributedWekaException

Update the underlying model for this scoring task

Parameters:

model - the new model to use

Throws:

DistributedWekaException - if the task has not yet been initialized by calling setModel().

processInstance

public double[] processInstance(weka.core.Instance inst)
                         throws DistributedWekaException

Process (score) an instance

Parameters:

inst - the instance to score

Returns:

a probability distribution over the labels that the model can predict

Throws:

DistributedWekaException - if a problem occurs

processInstanceBatchPredictor

public double[][] processInstanceBatchPredictor(weka.core.Instance inst)
                                         throws DistributedWekaException

Process an instance. When the batch size is met then a batch of predictions is returned; returns null if the batch size has not been matched yet

Parameters:

inst - the instance to process

Returns:

a batch of predictions or null if we have not seen enough input instances to meet the batch size yet

Throws:

DistributedWekaException - if a problem occurs

finalizeBatchPrediction

public double[][] finalizeBatchPrediction()
                                   throws DistributedWekaException

Finish off the last partial batch (if any).

Returns:

predictions for the last partial batch or null

Throws:

DistributedWekaException - if a problem occurs

getMissingMismatchAttributeInfo

public java.lang.String getMissingMismatchAttributeInfo()

Get a string summarizing missing and type mismatches between the incoming data and what the model expects

Returns:

a summary string of problems

getPredictionLabels

public java.util.List<java.lang.String> getPredictionLabels()
                                                     throws DistributedWekaException

Get a list of labels that the model can predict

Returns:

a list of labels that the model can predict

Throws:

DistributedWekaException - if a problem occurs

isBatchPredictor

public boolean isBatchPredictor()

Returns true if the underlying model is a BatchPredictor

Returns:

true if the underlying model is a BatchPredictor

modelIsUsingStringAttributes

public boolean modelIsUsingStringAttributes()

Returns true if model is using string attributes

Returns:

true if model is using string attributes

modelIsAClassifier

public boolean modelIsAClassifier()

Returns true if the underlying model is a classifier

Returns:

true if the underlying model is a classifier