Equidistant

Description and usage

This transformation is used to convert a equidistant time series to a non equidistant time series.

The transformation has two input time series. The first one is the equidistant time series which should be converted to a non-equidistant time series. The second time series is used as a reference time series. For each non-equidistant time step in this reference time series a non-equidistant time step will be created in the output time series. A single entry in the input time series can be re-used for sampling multiple entries in the output time series. For the equidistantInputVariable a non-equidistant time series can be supplied as well.

The value at this time step will derived from sampling the equidistant time series. It is possible to configure how the equidistant time series should be sampled.

The available options are:

closest
closestForward
closestBackward
linear
missing

Closest

For each time step in the output time series the closest value in the sampled equidistant input time series is used as the output value.

ClosestForward

For each time step in the output time series the closest value after the output time step in the sampled equidistant input time series is used as the output value.

ClosestBackward

For each time step in the output time series the closest value before the output time step in the sampled equidistant input time series is used as the output value.

Linear

For each time step in the output time series the output value will be determined by linear interpolation between the closest value before the output time step and the closest value

after the output time step

Missing

All output values are set to missing.

Input/Output time series

In this function two input time series and one output time series must be identified.

An equidistant input time series with an equidistant time step,
A non-equidistant input time series with the reference time steps for the output time series,
An output time series with the same time steps as the non-equidistant time series, but with the values sampled from the equidistant time series
Each time series is assigned a variable ID which is used in the actual expression. For definition of input and output time series see Transformation - Inputs and Outputs.

Configuration

A basic configuration of the function is described below. This describes the main elements and attributes required and provides an example configuration.

equidistantInputVariable

required element defined the equidistant time series which should be sampled.

timeReferenceInputVariable

required element defining the non-equidistant time steps for which output should be generated.

interpolationType

required element which defines which sampling method should be used.

Example

	<transformation id="Sample">
	<sample>
		<equidistant>
			<equidistantInputVariable>
				<variableId>inputTimeSeries</variableId>
			</equidistantInputVariable>
			<timeReferenceInputVariable>
				<variableId>timeReferenceTimeSeries</variableId>
			</timeReferenceInputVariable>
			<interpolationType>closest</interpolationType>
			<outputVariable>
				<variableId>output</variableId>
			</outputVariable>
		</equidistant>
	</sample>

Advanced configuration topics

Additional advanced options are available and are briefly described. See schema for information on how to configure advanced options.

className

Optional element to define a Java class to override the default class used for the transformation. If used this class must be available and implement the required transformation interface. See transformation - user defined classes for details

forecastLoopSearchPeriod

When a forecastLoopSearchPeriod is configured the transformation will be repeated for each external forecast found in the defined period.

This will only work when the <equidistantInputVariable> and <outputVariable> are external forecasts, the output variable for each execution will get the same external forecast time.

When <timeReferenceInputVariable> is also an external forecast, the same amount of external forecasts should be present as in the <equidistantInputVariable> because these will also be looped over.

When <timeReferenceInputVariable> is not an external forecast there will only be 1 time series available, this one be be reused for step of the loop.

Example forecastLoopSearchPeriod

	<transformation id="Sample">
	<sample>
		<equidistant>
			<equidistantInputVariable>
				<variableId>inputTimeSeries</variableId>
			</equidistantInputVariable>
			<timeReferenceInputVariable>
				<variableId>timeReferenceTimeSeries</variableId>
			</timeReferenceInputVariable>
			<forecastLoopSearchPeriod unit="week" start="-4" end="0"/>
			<interpolationType>closest</interpolationType>
			<outputVariable>
				<variableId>output</variableId>
			</outputVariable>
		</equidistant>
	</sample>

Page tree