Lag (Time Series)

Synopsis

This operator performs a time series lag transformation on one or more attributes.

Description

Different attributes can be lagged separately with different lag values by the parameter attributes. If the parameter overwrite attributes is selected, the lagged attributes overwrite the original ones. If not selected, new attributes are added to the ExampleSet (the names of the new attributes are in the form <attribute-name> - <lag>).

If the parameter extend exampleset is selected, the resulting ExampleSet is extended by n examples where n is the maximum lag specified. Attributes that are not selected for lagging are filled with missing values.

This operator works on all attributes (independent of type or role).

Input

  • example set input (IOObject)

    The ExampleSet which contains the time series data as attributes.

Output

  • example set output (IOObject)

    The ExampleSet after applying the lag transformation. If overwrite attributes is true, the original time series attributes are overwritten. Else new attributes with the lagged values are added. The names of the new attributes are in the form <attribute-name> - <lag>. If the parameter extend exampleset is selected, the resulting ExampleSet is extended by n examples where n is the maximum lag specified. Attributes that are not selected for lagging are filled with missing values.

  • original (IOObject)

    The ExampleSet that was given as input is passed through without changes.

Parameters

  • attributes

    The lag attributes can be selected by the drop down menu if the meta data is known. For each attribute an integer lag value has to be specified. If overwrite attributes is not selected, the same attribute can be lagged more than one time with different lag values.

    Range:
  • attribute

    The lag attribute can be selected by the drop down menu if the meta data is known. It can also be typed in manually.

    Range:
  • lag

    This parameter defines the number of lags for the attribute. Example i will contain the value of Example i-lag. The first lag values will be filled with missing values.

    Range:
  • overwrite_attributes

    This parameter indicates if the original time series attributes are overwritten by the lagged time series. If this parameter is set to false, the lagged time series are added as new attributes to the ExampleSet. The name of these new attributes will be <attribute-name> - <lag>.

    Note that selecting this parameter can increase runtime (it required copying the input ExampleSet to ensure that there are no data leaks).

    Range:
  • extend_exampleset

    This parameter indicates if the ExampleSet should be extended by n Examples (where n is the maximum lag specified). Attributes that are not selected for lagging are filled with missing values.

    Note that selecting this parameter can increase runtime (it required copying the input ExampleSet to ensure that there are no data leaks).

    Range:

Tutorial Processes

Lagging Lake Huron Data Set

In this tutorial process the lagging of the Lake Huron Data Set is demonstrated.

Lagging options demonstrated on the Golf data set

This tutorial process the different options for lagging are demonstrated on the Golf data set.