org.hawkular.datamining.forecast.models

Class DoubleExponentialSmoothing

java.lang.Object

org.hawkular.datamining.forecast.models.AbstractExponentialSmoothing

org.hawkular.datamining.forecast.models.DoubleExponentialSmoothing

All Implemented Interfaces:

TimeSeriesModel

public class DoubleExponentialSmoothing
extends AbstractExponentialSmoothing

Double exponential smoothing (Holt's linear trend model) Works well when data exhibits increasing or decreasing trend pattern.

When smoothing parameters are smaller more weights are added to the observations from distant past - it makes it model more robust.

Equations:

• level_t = alpha*y_t + (1-alpha)*(level_t-1+trend_t-1)
• trend_t = beta*(level_t-level_t-1) + (1-beta)*trend_t-1
• forecast_t+h = level_t + h*trend_t

Author:

Pavol Loffay

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	DoubleExponentialSmoothing.DoubleExOptimizer
static class	DoubleExponentialSmoothing.DoubleExState

Nested classes/interfaces inherited from class org.hawkular.datamining.forecast.models.AbstractExponentialSmoothing

AbstractExponentialSmoothing.PredictionResult

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

TimeSeriesModel.TimestampComparator

Field Summary

Fields

Modifier and Type	Field and Description
static double	DEFAULT_LEVEL_SMOOTHING
static double	DEFAULT_TREND_SMOOTHING
static double	MAX_LEVEL_TREND_SMOOTHING
static double	MIN_LEVEL_TREND_SMOOTHING

Fields inherited from class org.hawkular.datamining.forecast.models.AbstractExponentialSmoothing

lastTimestamp, metricContext

Method Summary

Modifier and Type	Method and Description
protected AbstractExponentialSmoothing.PredictionResult	calculatePrediction(int nAhead, Long learnTimestamp, Double expected)
static DoubleExponentialSmoothing	createCustom(double levelSmoothing, double trendSmoothing, MetricContext metricContext)
static DoubleExponentialSmoothing	createDefault()
static DoubleExponentialSmoothing	createWithMetric(MetricContext metricContext)
static DoubleExponentialSmoothing	createWithSmoothingParams(double levelSmoothing, double trendSmoothing)
protected DoubleExponentialSmoothing.DoubleExState	initState(List<DataPoint> initData)
int	minimumInitSize()
String	name()
int	numberOfParams()
static DoubleExponentialSmoothing.DoubleExOptimizer	optimizer()
static DoubleExponentialSmoothing.DoubleExOptimizer	optimizer(MetricContext metricContext)
protected DoubleExponentialSmoothing.DoubleExState	state()
String	toString()
protected void	updateState(DataPoint dataPoint)

Methods inherited from class org.hawkular.datamining.forecast.models.AbstractExponentialSmoothing

forecast, forecast, init, initStatistics, lastTimestamp, learn, learn, runStatistics, setConfidenceInterval

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_LEVEL_SMOOTHING

public static final double DEFAULT_LEVEL_SMOOTHING

See Also:

Constant Field Values

DEFAULT_TREND_SMOOTHING

public static final double DEFAULT_TREND_SMOOTHING

See Also:

Constant Field Values

MIN_LEVEL_TREND_SMOOTHING

public static final double MIN_LEVEL_TREND_SMOOTHING

See Also:

Constant Field Values

MAX_LEVEL_TREND_SMOOTHING

public static final double MAX_LEVEL_TREND_SMOOTHING

See Also:

Constant Field Values

Method Detail

createDefault

public static DoubleExponentialSmoothing createDefault()

createWithMetric

public static DoubleExponentialSmoothing createWithMetric(MetricContext metricContext)

createWithSmoothingParams

public static DoubleExponentialSmoothing createWithSmoothingParams(double levelSmoothing,
                                                                   double trendSmoothing)

createCustom

public static DoubleExponentialSmoothing createCustom(double levelSmoothing,
                                                      double trendSmoothing,
                                                      MetricContext metricContext)

name

public String name()

Returns:

name of the implemented model

numberOfParams

public int numberOfParams()

Returns:

number of parameters of the model, seasonal models include seasonal indices

minimumInitSize

public int minimumInitSize()

Returns:

number of parameters of the model, seasonal models include seasonal indices

initState

protected DoubleExponentialSmoothing.DoubleExState initState(List<DataPoint> initData)

Specified by:

initState in class AbstractExponentialSmoothing

state

protected DoubleExponentialSmoothing.DoubleExState state()

Specified by:

state in class AbstractExponentialSmoothing

updateState

protected void updateState(DataPoint dataPoint)

Specified by:

updateState in class AbstractExponentialSmoothing

calculatePrediction

protected AbstractExponentialSmoothing.PredictionResult calculatePrediction(int nAhead,
                                                                            Long learnTimestamp,
                                                                            Double expected)

Specified by:

calculatePrediction in class AbstractExponentialSmoothing

optimizer

public static DoubleExponentialSmoothing.DoubleExOptimizer optimizer()

optimizer

public static DoubleExponentialSmoothing.DoubleExOptimizer optimizer(MetricContext metricContext)

toString

public String toString()

Overrides:

toString in class Object