Package org.processmining.plugins.stochasticpetrinet.simulator.timeseries

Class TimeSeries<H>

java.lang.Object

org.processmining.plugins.stochasticpetrinet.simulator.timeseries.TimeSeries<H>

Type Parameters:

H - the predicted variable which is supposedly correlated to its previous values.

Direct Known Subclasses:

AverageMethodTimeSeries, DriftMethodTimeSeries, LastObservationTimeSeries, NaiveMethodTimeSeries, RTimeSeries, SeasonalNaiveMethodTimeSeries

public abstract class TimeSeries<H>
extends java.lang.Object

A time series of values (might be durations, might be rates, might be anything)

Author:

Andreas Rogge-Solti

Field Summary

Fields

Modifier and Type	Field	Description
protected java.lang.String	key	A unique key for this time series
protected int	lag	the window to use for predictions
protected int	season	the seasonality of the data (is there seasonality after 7 days?)

Constructor Summary

Constructors

Constructor	Description
TimeSeries()

Method Summary

Modifier and Type	Method	Description
void	addObservation(Observation<H> observation)
protected Observation<H>	findLastAvailableObservation()
protected abstract void	fit(LimitedQueue<Observation<H>> currentObservations)	Subclasses need to fit their corresponding time series model to the currently available data.
Observation<H>	getLastObservation()
Observation<H>	getObservationOfLastSeason(long index)
protected abstract Prediction<H>	getPrediction(int h, java.lang.Object... payload)	Subclasses need to provide an implementation for the prediction of h-steps into the future given the current position
protected abstract boolean	isAvailable(H observation)
Prediction<H>	predict(int h, java.lang.Object... payload)	Performs a forecast based on the current observations h steps ahead.
void	resetTo(java.util.List<Observation<H>> observations)	resets the current observations to the passed observations (useful for initialization)
void	setKey(java.lang.String key)

Methods inherited from class java.lang.Object

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

Field Detail

lag

protected int lag

the window to use for predictions

season

protected int season

the seasonality of the data (is there seasonality after 7 days?)

key

protected java.lang.String key

A unique key for this time series

Constructor Detail

TimeSeries

public TimeSeries()

Method Detail

resetTo

public void resetTo(java.util.List<Observation<H>> observations)

resets the current observations to the passed observations (useful for initialization)

Parameters:

observations - List of Observations

predict

public Prediction<H> predict(int h,
                             java.lang.Object... payload)

Performs a forecast based on the current observations h steps ahead.

Parameters:

h - the forecast horizon

Returns:

Prediction a prediction

addObservation

public void addObservation(Observation<H> observation)

getLastObservation

public Observation<H> getLastObservation()

findLastAvailableObservation

protected Observation<H> findLastAvailableObservation()

getObservationOfLastSeason

public Observation<H> getObservationOfLastSeason(long index)

isAvailable

protected abstract boolean isAvailable(H observation)

setKey

public void setKey(java.lang.String key)

fit

protected abstract void fit(LimitedQueue<Observation<H>> currentObservations)

Subclasses need to fit their corresponding time series model to the currently available data.

getPrediction

protected abstract Prediction<H> getPrediction(int h,
                                               java.lang.Object... payload)

Subclasses need to provide an implementation for the prediction of h-steps into the future given the current position

Parameters:

h - the prediction/forecast horizon

payload - additional information which is available at time of prediction (if there are other additional explanatory variables)