Search

Using Auto Regressive to solve Regression Problems

The Auto Regressive task predicts the future values assumed by a signal over time, principally based on its past behavior.

Prerequisites

you must have created a flow;
the required datasets must have been imported into the flow;
the data used for the analysis must have been well prepared;
the time series and exogenous inputs are non-categorical.

Additional tabs

The Models tab, where the coefficient values are displayed, along with any periods that have been retrieved, and other statistical parameters, such as minimum and maximum.
The Results tab, where statistics on the AR computation are displayed, such as the execution time.

Procedure

Drag and drop the Auto Regressive task onto the stage.
Connect a task, which contains the attributes from which you want to create the model, to the new task.
Double click the Auto Regressive task. The left-hand pane displays a list of all the available attributes in the dataset, which can be ordered and searched as required.
Configure the options described in the table below.
Save and compute the task.

Parameter Name	Description
Auto Regressive basic options
Time series attributes (ORDERED)	The ordered attribute which will be used to perform auto regression.
Exogenous attributes (ORDERED)	Additional attributes that need to be considered when performing auto regression predictions. For example, the external temperature could be considered an exogenous attribute when calculating the consumption of energy in heating a house.
ID Attributes (NOMINAL)	The nominal key attributes that define the rows that will contain the values to be used. Several key attributes can be used: each different tuple of values assumed by the key variables identifies a set of rows concerning a separate model. Consequently the task builds as many models as the different tuples of the key variables. Drag and drop here the key attributes. Rulex will create a different sequence for each key attribute value. Instead of manually dragging and dropping attributes, they can be defined via a filtered list.
Interval dimension for time series attributes	The number of previous values to be used in the calculation for the time series attribute.
Interval dimension for exogenous attributes	The number of previous values to be used in the calculation for each exogenous attribute.
Time attribute	Select a Time attribute from the drop-down list to specify the temporal variable to be used. If not specified, row numbers will be used to establish the time relationship.
Time lag	The length of each time interval.
Time unit	Select the unit of time required. Possible values are: Second(s) Minute(s) Hour(s) Day(s) Week(s) Month(s) Quarter(s) Year(s)

Parameter Name	Description
Auto Regressive advanced options
Percentage of constant values for switching to Croston model	When the percentage specified is exceeded the Croston model is used instead of the auto regressive model in the calculation.
Percentage tolerance for switching to Croston model	The tolerance percentage in the range of variation for constant values. When this percentage is not exceeded, the values are considered constant when deciding if the Croston model is to be used or not.
Obsolescence coefficients	The column that contains the obsolescence coefficient to be applied to the time series.
Smoothing function on time series	The function to be used to smooth the values in the time series. Possible values are: Log (transforms the input variables using the log-exponential smoothing technique) Box-Cox (transforms the input variable using a Box-Cox transformation) None (no function is used to smooth the input variable)
Interval dimension on time series for rolling mean	If you want to use the rolling mean of past values instead of single values, specify here how many values will be contained in each rolling value group. The number of groups used is specified in the Interval dimension option.