com.github.signaflo.timeseries.model.arima

Interface Arima

All Superinterfaces:

Model

public interface Arima
extends Model

A seasonal autoregressive integrated moving average (ARIMA) model.

Nested Class Summary

Nested Classes

Modifier and Type	Interface and Description
static class	Arima.Constant An indicator for whether an ARIMA model has a constant term.
static class	Arima.Drift An indicator for whether an ARIMA model has a drift term.
static class	Arima.FittingStrategy The strategy to be used for fitting an ARIMA model.

Method Summary

Modifier and Type	Method and Description
double	aic() Get the Akaike Information Criterion (AIC) for this model.
ArimaCoefficients	coefficients() Get the coefficients of this ARIMA model.
TimeSeries	fittedSeries() Get the model fitted values, which are in-sample one-step ahead forecasts.
Forecast	forecast(int steps, double alpha) Produce a forecast from this model up to the given number of steps ahead.
double	logLikelihood() Get the natural logarithm of the likelihood of the model given the data.
static Arima	model(TimeSeries observations, ArimaCoefficients coeffs, Arima.FittingStrategy fittingStrategy) Create a new ARIMA model from the given observations, model coefficients, and fitting strategy.
static Arima	model(TimeSeries observations, ArimaCoefficients coeffs, TimePeriod seasonalCycle) Create a new ARIMA model from the given observations, model coefficients, and seasonal cycle.
static Arima	model(TimeSeries observations, ArimaCoefficients coeffs, TimePeriod seasonalCycle, Arima.FittingStrategy fittingStrategy) Create a new ARIMA model from the given observations, model coefficients, seasonal cycle, and fitting strategy.
static Arima	model(TimeSeries observations, ArimaOrder order) Create a new ARIMA model from the given observations and model order.
static Arima	model(TimeSeries observations, ArimaOrder order, Arima.FittingStrategy fittingStrategy) Create a new ARIMA model from the given observations, model order, and fitting strategy.
static Arima	model(TimeSeries observations, ArimaOrder order, TimePeriod seasonalCycle) Create a new ARIMA model from the given observations, model order, and seasonal cycle.
static Arima	model(TimeSeries observations, ArimaOrder order, TimePeriod seasonalCycle, Arima.FittingStrategy fittingStrategy) Create a new ARIMA model from the given observations, model order, seasonal cycle, and fitting strategy.
TimeSeries	observations() Get the series of observations.
ArimaOrder	order() Get the order of this ARIMA model.
TimeSeries	predictionErrors() Get the model prediction errors, the difference between the observed values and the model fitted values.
int	seasonalFrequency() Get the frequency of observations per seasonal cycle.
double	sigma2() Get the estimate of the process variance.
double[]	stdErrors() Get the standard errors of the model coefficients.

Methods inherited from interface com.github.signaflo.timeseries.model.Model

forecast

Method Detail

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaOrder order)

Create a new ARIMA model from the given observations and model order. This constructor sets the model Arima.FittingStrategy to conditional sum-of-squares with maximum likelihood and the seasonal cycle to one year.

Parameters:

observations - the time series of observations.

order - the order of the ARIMA model.

Returns:

a new ARIMA model from the given observations and model order.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaOrder order,
                   @NonNull
                   TimePeriod seasonalCycle)

Create a new ARIMA model from the given observations, model order, and seasonal cycle. This method sets the model Arima.FittingStrategy to conditional sum-of-squares with maximum likelihood.

Parameters:

observations - the time series of observations.

order - the order of the ARIMA model.

seasonalCycle - the amount of time it takes for the seasonal pattern to complete one cycle. For example, monthly data usually has a cycle of one year, hourly data a cycle of one day, etc... However, a seasonal cycle may be an arbitrary amount of time.

Returns:

a new ARIMA model from the given observations, model order, and seasonal cycle.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaOrder order,
                   @NonNull
                   Arima.FittingStrategy fittingStrategy)

Create a new ARIMA model from the given observations, model order, and fitting strategy. This method sets the seasonal cycle to one year.

Parameters:

observations - the time series of observations.

order - the order of the ARIMA model.

fittingStrategy - the strategy to use to fit the model to the data. Maximum-likelihood estimates are typically preferred for greater precision and accuracy, but take longer to obtain than conditional sum-of-squares estimates.

Returns:

a new ARIMA model from the given observations, model order, and fitting strategy.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaOrder order,
                   @NonNull
                   TimePeriod seasonalCycle,
                   @NonNull
                   Arima.FittingStrategy fittingStrategy)

Create a new ARIMA model from the given observations, model order, seasonal cycle, and fitting strategy.

Parameters:

observations - the time series of observations.

order - the order of the ARIMA model.

seasonalCycle - the amount of time it takes for the seasonal pattern to complete one cycle. For example, monthly data usually has a cycle of one year, hourly data a cycle of one day, etc... However, a seasonal cycle may be an arbitrary amount of time.

fittingStrategy - the strategy to use to fit the model to the data. Maximum-likelihood estimates are typically preferred for greater precision and accuracy, but take longer to obtain than conditional sum-of-squares estimates.

Returns:

a new ARIMA model from the given observations, model order, seasonal cycle, and fitting strategy.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaCoefficients coeffs,
                   @NonNull
                   Arima.FittingStrategy fittingStrategy)

Create a new ARIMA model from the given observations, model coefficients, and fitting strategy. This constructor sets the seasonal cycle to one year.

Parameters:

observations - the time series of observations.

coeffs - the coefficients of the model.

fittingStrategy - the strategy to use to fit the model to the data. Maximum-likelihood estimates are typically preferred for greater precision and accuracy, but take longer to obtain than conditional sum-of-squares estimates.

Returns:

a new ARIMA model from the given observations, model coefficients, and fitting strategy.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaCoefficients coeffs,
                   @NonNull
                   TimePeriod seasonalCycle)

Create a new ARIMA model from the given observations, model coefficients, and seasonal cycle. This constructor sets the model Arima.FittingStrategy to conditional sum-of-squares with maximum likelihood.

Parameters:

observations - the time series of observations.

coeffs - the coefficients of the model.

seasonalCycle - the amount of time it takes for the seasonal pattern to complete one cycle. For example, monthly data usually has a cycle of one year, hourly data a cycle of one day, etc... However, a seasonal cycle may be an arbitrary amount of time.

Returns:

a new ARIMA model from the given observations, model coefficients, and seasonal cycle.

model

static Arima model(@NonNull
                   TimeSeries observations,
                   @NonNull
                   ArimaCoefficients coeffs,
                   @NonNull
                   TimePeriod seasonalCycle,
                   @NonNull
                   Arima.FittingStrategy fittingStrategy)

Create a new ARIMA model from the given observations, model coefficients, seasonal cycle, and fitting strategy.

Parameters:

observations - the time series of observations.

coeffs - the coefficients of the model.

seasonalCycle - the amount of time it takes for the seasonal pattern to complete one cycle. For example, monthly data usually has a cycle of one year, hourly data a cycle of one day, etc... However, a seasonal cycle may be an arbitrary amount of time.

fittingStrategy - the strategy to use to fit the model to the data. Maximum-likelihood estimates are typically preferred for greater precision and accuracy, but take longer to obtain than conditional sum-of-squares estimates.

Returns:

a new ARIMA model from the given observations, model coefficients, seasonal cycle, and fitting strategy.

forecast

Forecast forecast(int steps,
                  double alpha)

Description copied from interface: Model

Produce a forecast from this model up to the given number of steps ahead.

Specified by:

forecast in interface Model

Parameters:

steps - the number of time periods ahead to forecast.

alpha - the significance level to use for the prediction interval.

Returns:

a forecast from this model up to the given number of steps ahead.

observations

TimeSeries observations()

Description copied from interface: Model

Get the series of observations.

Specified by:

observations in interface Model

Returns:

the series of observations.

fittedSeries

TimeSeries fittedSeries()

Description copied from interface: Model

Get the model fitted values, which are in-sample one-step ahead forecasts.

Specified by:

fittedSeries in interface Model

Returns:

the model fitted values.

predictionErrors

TimeSeries predictionErrors()

Description copied from interface: Model

Get the model prediction errors, the difference between the observed values and the model fitted values.

Specified by:

predictionErrors in interface Model

Returns:

the model prediction errors.

sigma2

double sigma2()

Get the estimate of the process variance.

Returns:

the estimate of the process variance.

seasonalFrequency

int seasonalFrequency()

Get the frequency of observations per seasonal cycle.

Returns:

the frequency of observations per seasonal cycle.

stdErrors

double[] stdErrors()

Get the standard errors of the model coefficients.

Returns:

the standard errors of the model coefficients.

coefficients

ArimaCoefficients coefficients()

Get the coefficients of this ARIMA model.

Returns:

the coefficients of this ARIMA model.

order

ArimaOrder order()

Get the order of this ARIMA model.

Returns:

the order of this ARIMA model.

logLikelihood

double logLikelihood()

Get the natural logarithm of the likelihood of the model given the data.

Returns:

the natural logarithm of the likelihood of the model given the data.

aic

double aic()

Get the Akaike Information Criterion (AIC) for this model. The AIC is defined as 2k − 2L where k is the number of parameters in the model and L is the logarithm of the likelihood.

Returns:

the Akaike Information Criterion (AIC) for this model.