org.hawkular.datamining.forecast.models

Class AbstractExponentialSmoothing

java.lang.Object

org.hawkular.datamining.forecast.models.AbstractExponentialSmoothing

All Implemented Interfaces:

TimeSeriesModel

Direct Known Subclasses:

DoubleExponentialSmoothing, SimpleExponentialSmoothing, TripleExponentialSmoothing

public abstract class AbstractExponentialSmoothing
extends Object
implements TimeSeriesModel

Author:

Pavol Loffay

Nested Class Summary

Nested classes/interfaces inherited from interface org.hawkular.datamining.forecast.models.TimeSeriesModel

TimeSeriesModel.TimestampComparator

Field Summary

Fields

Modifier and Type	Field and Description
protected long	lastTimestamp
protected MetricContext	metricContext

Constructor Summary

Constructors

Constructor and Description
AbstractExponentialSmoothing(MetricContext metricContext)

Method Summary

Modifier and Type	Method and Description
protected abstract double	calculatePrediction(long nAhead, Long learnTimestamp)
DataPoint	forecast() One step ahead prediction
List<DataPoint>	forecast(int nAhead) Multi step ahead prediction
AccuracyStatistics	init(List<DataPoint> dataPoints) Initialize model and return statistics - error of one step ahead prediction
protected abstract SimpleExponentialSmoothing.State	initState(List<DataPoint> dataPoints)
AccuracyStatistics	initStatistics()
long	lastTimestamp()
void	learn(DataPoint dataPoint) Learn one point
void	learn(List<DataPoint> dataPoints) Learn multiple points
AccuracyStatistics	runStatistics()
protected abstract SimpleExponentialSmoothing.State	state()
protected abstract void	updateState(DataPoint dataPoint)

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.hawkular.datamining.forecast.models.TimeSeriesModel

minimumInitSize, name, numberOfParams

Field Detail

lastTimestamp

protected long lastTimestamp

metricContext

protected MetricContext metricContext

Constructor Detail

AbstractExponentialSmoothing

public AbstractExponentialSmoothing(MetricContext metricContext)

Method Detail

calculatePrediction

protected abstract double calculatePrediction(long nAhead,
                                              Long learnTimestamp)

updateState

protected abstract void updateState(DataPoint dataPoint)

initState

protected abstract SimpleExponentialSmoothing.State initState(List<DataPoint> dataPoints)

state

protected abstract SimpleExponentialSmoothing.State state()

init

public AccuracyStatistics init(List<DataPoint> dataPoints)

Description copied from interface: TimeSeriesModel

Initialize model and return statistics - error of one step ahead prediction

Specified by:

init in interface TimeSeriesModel

Parameters:

dataPoints - points to learn

Returns:

statistics for learnData

learn

public void learn(DataPoint dataPoint)

Description copied from interface: TimeSeriesModel

Learn one point

Specified by:

learn in interface TimeSeriesModel

Parameters:

dataPoint - point to learn

learn

public void learn(List<DataPoint> dataPoints)

Description copied from interface: TimeSeriesModel

Learn multiple points

Specified by:

learn in interface TimeSeriesModel

Parameters:

dataPoints - points to learn

forecast

public DataPoint forecast()

Description copied from interface: TimeSeriesModel

One step ahead prediction

Specified by:

forecast in interface TimeSeriesModel

Returns:

predicted point

forecast

public List<DataPoint> forecast(int nAhead)

Description copied from interface: TimeSeriesModel

Multi step ahead prediction

Specified by:

forecast in interface TimeSeriesModel

Parameters:

nAhead - number of steps for forecasting

Returns:

predicted points

initStatistics

public AccuracyStatistics initStatistics()

Specified by:

initStatistics in interface TimeSeriesModel

Returns:

initial statistics calculated when calling TimeSeriesModel.initStatistics()

runStatistics

public AccuracyStatistics runStatistics()

Specified by:

runStatistics in interface TimeSeriesModel

Returns:

statistics calculated on learning data

lastTimestamp

public long lastTimestamp()

Specified by:

lastTimestamp in interface TimeSeriesModel