org.hawkular.datamining.forecast

Class AutomaticForecaster

java.lang.Object

org.hawkular.datamining.forecast.AutomaticForecaster

All Implemented Interfaces:

Forecaster

public class AutomaticForecaster
extends Object
implements Forecaster

Selects the best model for given data set.

Author:

Pavol Loffay

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	AutomaticForecaster.ConceptDriftStrategy Strategy used for dealing with concept drift (statistical properties of modelled time series change over time) For instance originally monotonic time series has changed to trend stationary or seasonal pattern has showed up.
static class	AutomaticForecaster.ErrorChangeStrategy Changes model if run accuracy differs by X percent to init accuracy (MSE, MAE).
static class	AutomaticForecaster.PeriodicIntervalStrategy The best model is periodically selected after each x learned points.

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

Forecaster.Config, Forecaster.Update

Constructor Summary

Constructors

Constructor and Description
AutomaticForecaster(MetricContext context)
AutomaticForecaster(MetricContext context, Forecaster.Config config)

Method Summary

Modifier and Type	Method and Description
Forecaster.Config	config()
MetricContext	context()
DataPoint	forecast()
List<DataPoint>	forecast(int nAhead)
boolean	initialized()
long	lastTimestamp()
void	learn(DataPoint dataPoint)
void	learn(List<DataPoint> dataPoints)
TimeSeriesModel	model()
void	update(Forecaster.Update update)

Methods inherited from class java.lang.Object

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

Constructor Detail

AutomaticForecaster

public AutomaticForecaster(MetricContext context)

AutomaticForecaster

public AutomaticForecaster(MetricContext context,
                           Forecaster.Config config)

Method Detail

learn

public void learn(DataPoint dataPoint)

Specified by:

learn in interface Forecaster

See Also:

TimeSeriesModel.learn(DataPoint)

learn

public void learn(List<DataPoint> dataPoints)

Specified by:

learn in interface Forecaster

See Also:

TimeSeriesModel.learn(List)

forecast

public DataPoint forecast()

Specified by:

forecast in interface Forecaster

See Also:

TimeSeriesModel.forecast()

forecast

public List<DataPoint> forecast(int nAhead)

Specified by:

forecast in interface Forecaster

See Also:

TimeSeriesModel.forecast(int)

model

public TimeSeriesModel model()

Specified by:

model in interface Forecaster

Returns:

currently used model

context

public MetricContext context()

Specified by:

context in interface Forecaster

Returns:

information about metric

initialized

public boolean initialized()

Specified by:

initialized in interface Forecaster

Returns:

true if model is selected is initialized

lastTimestamp

public long lastTimestamp()

Specified by:

lastTimestamp in interface Forecaster

Returns:

last learned timestamp

update

public void update(Forecaster.Update update)

Specified by:

update in interface Forecaster

config

public Forecaster.Config config()

Specified by:

config in interface Forecaster