ARIMA is short for Auto Regressive Integrated Moving Average. This acronym is descriptive, capturing the key aspects of the model itself:

AR: Autoregression.
This part of the model expects that the time series has a dependent relationship between an observation and some number of lagged observations. In other words: The evolving observation value is regressed on its own past values.
I Integrated.
This part of the model makes it applicable to use cases where data show evidence of non-stationarity. Initial differencing steps can be used one or more times to eliminate the non-stationarity from raw observations.
MA Moving Average.
This model part indicates that the regression error is actually a linear combination of error terms whose values occurred contemporaneously and at various times in the past. A model that uses the dependency between an observation and a residual error from a moving average model applied to lagged observations.

An ARIMA model is a generalization of an AutoRegressive Moving Average (ARMA) model. Both models are fitted to time series data to forecast future points in the series.

Plugins

TsARIMASink

@Plugin(type = SparkSink.PLUGIN_TYPE)
@Name("TsARIMASink")
@Description("A building stage for an Apache Spark based ARIMA model for time series datasets.")
public class TsARIMASink extends ARIMASink {

    ...

}

Parameters

Model Name The unique name of the ARIMA model.
Time Field The name of the field in the input schema that contains the time value.
Value Field The name of the field in the input schema that contains the value.
Time Split The split of the dataset into train & test data, e.g. 80:20. Note, this is a split time and is computed from the total time span (min, max) of the time series. Default is 70:30."
Model Configuration
Lag Order The positive number of lag observations included in the time series model (also called the lag order).
Differencing Degree The positive number of times that the raw observations are differenced (also called the degree of differencing).
Moving Average Order A positive number that specifies the size of the moving average window (also called the order of moving average).
ElasticNet Mixing The ElasticNet mxing parameter. For value = 0.0, the penalty is an L2 penalty. For value = 1.0, it is an L1 penalty. For 0.0 < value < 1.0, the penalty is a combination of L1 and L2. Default is 0.0.
Regularization Parameter The nonnegative regularization parameter. Default is 0.0.
Standardization The indicator to determine whether to standardize the training features before fitting the model. Default is 'true'."
With Intercept The indicator to determine whether to fit an intercept value.
Remove Mean The indicator to determine whether to remove the mean value from the value from the value of the time series before training model. Default is 'false'.

TsARIMA

@Plugin(type = SparkCompute.PLUGIN_TYPE)
@Name("TsARIMA")
@Description("A transformation stage that leverages a trained ARIMA model to look n steps in time ahead. "
		+ "The forecast result is described by a two column output schema, one column specifies the future "
		+ "points in time, and another the forecasted values.")
public class TsARIMA extends ARIMACompute {

    ...

}

Parameters

Model Name The unique name of the ARIMA model.
Time Field The name of the field in the input schema that contains the time value.
Value Field The name of the field in the input schema that contains the value.
Time Steps The positive number of discrete time steps to look ahead. Default is 1.

TsAutoARIMASink

@Plugin(type = SparkSink.PLUGIN_TYPE)
@Name("TsAutoARIMASink")
@Description("A building stage for an Apache Spark based AutoARIMA model for time series datasets.")
public class TsAutoARIMASink extends ARIMASink {

    ...

}

Parameters

Model Name The unique name of the AutoARIMA model.
Time Field The name of the field in the input schema that contains the time value.
Value Field The name of the field in the input schema that contains the value.
Time Split The split of the dataset into train & test data, e.g. 80:20. Note, this is a split time and is computed from the total time span (min, max) of the time series. Default is 70:30."
Model Configuration
Maximum Lag Order The positive upper limit for tuning the number of lag operations (p).
Maximum Differencing Degree The positive upper limit for tuning the degree of differencing (d).
Maximum Moving Average Order The positive upper limit for tuning the size of the moving average window (q).
ElasticNet Mixing The ElasticNet mxing parameter. For value = 0.0, the penalty is an L2 penalty. For value = 1.0, it is an L1 penalty. For 0.0 < value < 1.0, the penalty is a combination of L1 and L2. Default is 0.0.
Regularization Parameter The nonnegative regularization parameter. Default is 0.0.
Standardization The indicator to determine whether to standardize the training features before fitting the model. Default is 'true'."
With Intercept The indicator to determine whether to fit an intercept value.
Remove Mean The indicator to determine whether to remove the mean value from the value from the value of the time series before training model. Default is 'false'.
Info Criterion The information criterion to calculate for model parameter tuning. Supported values are 'aic' (Akaike Information Criterion), 'aicc' (AIC with correction for finite sample sizes) and 'bic' (Bayesian Information Criterion). Default is 'aic'.

TsAutoARIMA

@Plugin(type = SparkCompute.PLUGIN_TYPE)
@Name("TsAutoARIMA")
@Description("A transformation stage that leverages a trained AutoARIMA model to look n steps in time ahead. "
		+ "The forecast result is described by a two column output schema, one column specifies the future "
		+ "points in time, and another the forecasted values.")
public class TsAutoARIMA extends ARIMACompute {

    ...

}

Parameters

Model Name The unique name of the AutoARIMA model.
Time Field The name of the field in the input schema that contains the time value.
Value Field The name of the field in the input schema that contains the value.
Time Steps The positive number of discrete time steps to look ahead. Default is 1.