weka.classifiers.timeseries.core

Class ErrorBasedConfidenceIntervalEstimator

java.lang.Object

weka.classifiers.timeseries.core.ErrorBasedConfidenceIntervalEstimator

All Implemented Interfaces:

Serializable

public class ErrorBasedConfidenceIntervalEstimator
extends Object
implements Serializable

Class that computes confidence intervals for a time series forecaster using errors computed on the training data.

Version:

$Revision: 50211 $

Author:

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

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
ErrorBasedConfidenceIntervalEstimator()

Method Summary

Methods

Modifier and Type	Method and Description
void	calculateConfidenceOffsets(TSForecaster forecaster, weka.core.Instances insts, int numPrime, int numSteps, double confidenceLevel, PrintStream... progress) Computes confidence intervals using the supplied forecster and training data.
void	calculateConfidenceOffsets(TSForecaster forecaster, weka.core.Instances insts, int numPrime, int artificialTimeStartValue, int numSteps, double confidenceLevel, PrintStream... progress) Computes confidence intervals using the supplied forecaster and training data.
double	getConfidenceLevel() Get the confidence level in use
double[]	getConfidenceLimitsForTarget(String targetName, double targetValue, int stepNum) Get the confidence limits (upper and lower bounds) for the named target at the given step number
List<double[]>	getConfidenceOffsets(double confidenceLevel, List<List<weka.classifiers.evaluation.NumericPrediction>> predictions) Get the confidence bound offsets for each target at the supplied confidence level

Methods inherited from class java.lang.Object

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

Constructor Detail

ErrorBasedConfidenceIntervalEstimator

public ErrorBasedConfidenceIntervalEstimator()

Method Detail

getConfidenceOffsets

public List<double[]> getConfidenceOffsets(double confidenceLevel,
                                  List<List<weka.classifiers.evaluation.NumericPrediction>> predictions)
                                    throws Exception

Get the confidence bound offsets for each target at the supplied confidence level

Parameters:

confidenceLevel - the confidence level to use

Returns:

a List of confidence offsets - one for each target. Each target's confidence limits are stored in a two element array, where the first element stores the upper bound and the second the lower bound (both are expressed as an offset)

Throws:

Exception - if the confidence limits can't be computed for some reason

calculateConfidenceOffsets

public void calculateConfidenceOffsets(TSForecaster forecaster,
                              weka.core.Instances insts,
                              int numPrime,
                              int numSteps,
                              double confidenceLevel,
                              PrintStream... progress)
                                throws Exception

Computes confidence intervals using the supplied forecster and training data.

Parameters:

forecaster - the forecaster to use

insts - the training data to use

numPrime - the number of instances to prime the forecaster with

numSteps - the number of steps to forecast (and hence compute intervals for)

confidenceLevel - the confidence level to use

progress - PrintStream objects to report progress to

Throws:

Exception - if a problem occurs

calculateConfidenceOffsets

public void calculateConfidenceOffsets(TSForecaster forecaster,
                              weka.core.Instances insts,
                              int numPrime,
                              int artificialTimeStartValue,
                              int numSteps,
                              double confidenceLevel,
                              PrintStream... progress)
                                throws Exception

Computes confidence intervals using the supplied forecaster and training data.

Parameters:

forecaster - the forecaster to use

insts - the training data

numPrime - the number of instances to prime with

artificialTimeStartValue - start value for the artificial time stamp (if one is being used or -1 otherwise)

numSteps - number of time steps to compute confidence intervals for

confidenceLevel - the confidence level to use

progress - varargs PrintStream object(s) to report progress to

Throws:

Exception - if something goes wrong.

getConfidenceLevel

public double getConfidenceLevel()

Get the confidence level in use

Returns:

the confidence level

getConfidenceLimitsForTarget

public double[] getConfidenceLimitsForTarget(String targetName,
                                    double targetValue,
                                    int stepNum)
                                      throws Exception

Get the confidence limits (upper and lower bounds) for the named target at the given step number

Parameters:

targetName - the name of the target to return the limits for

targetValue - the predicted target value

stepNum - the step number to return the bounds for this target

Returns:

an array containing the lower and upper bounds for the supplied target value in elements 0 and 1 respectively.

Throws:

Exception - if a problem occurs while computing the bounds.