The tool supports interface of Foresight Analytics Platform 9 or earlier.
This method executes general non-linear data transformations.
The Specification panel for this method looks as follows:
By default, an output variable is not additionally transformed before model calculation.
To set additional transformation of output variable before model calculation, use the Output Variable Transformation drop-down list.
Use the Formula Actual Period drop-down list and specify the period, in which calculation is executed:
Sample Period. The formula is applied to the sample period values. The period is set on the Calculation Periods panel.
Forecasting Period. The formula is applied to values of the forecasting period. The period is set on the Calculation Periods panel.
Whole Period. The formula is applied to the values of the sample and forecasting period.
NOTE. It is available to set formula actual period, if data source of output variable contains calendar dimension.
Use the Calculation Mode drop-down list. The setup is available if modeling and forecasting is working in attribute mode, and output variable data source contains calendar dimension. Select the mode to calculate the expression:
Series. It is used by default. The expression is calculated for all series points at the same time.
Pointwise (Forward). The expression is calculated sequentially in each point of the variable: first, the expression is calculated by the first point, then by the second point, and so on.
Pointwise (Backward). The expression is calculated sequentially in each point of the variable: first, the expression is calculated by its last point, then by the point before last, and so on.
NOTE. This setting is used only to calculate transformation problems.
To determine a set of explanatory variables (factors) that affect the output variable, use the Factors list. An element with the identifier like Xn, where n is the sequential number is assigned to each variable. These elements are used when setting up the equation.
Basic principles of working with factors are given in the Working with Factors (Input Variables) section.
The model enables the user to set custom variable frequency. To do this, select the Frequency item in the variable's menu and in the submenu that opens specify the required frequency. By default, the variable frequency matches the model frequency.
In order the model frequency variable is used in the context menu of the variable, select the Frequency item and select the Model Frequency radio button in the submenu that opens.
NOTE. Variable frequency can be changed only for variables, which data source contains a calendar dimension.
To use variable in the equation
Select a variable and click the Insert into Equation button.
Select the Insert into Equation item in the variable's context menu.
Drag the variable from the Factors list.
Enter identifier of the element corresponding to the variable.
The element that corresponds to the selected variable is inserted into equation.
To compose a model equation, use the Y= box. The data transformation equation is created in this box. All elements corresponding to variables are shown as hyperlinks in the equation. Hovering over a hyperlink shows a tooltip containing the name of the corresponding variable. In the equation, you can use any arithmetic operations, parentheses, statistical, mathematical and custom functions or methods.
Click the Open in Editor button to edit the equation. In the Expression Editor dialog box that opens make the required changes.
NOTE. The Y= box name is changed according to transformation set up for the output variable. For example, if the Natural Logarithm transformation is selected, the box has Log(Y)= name.
Select the Forecast Correction checkbox.
Set a variable used for forecast correction. This variable is not included into the identified equation of the model.
Forecast correction is applied in the model.
On working of modeling and forecasting in the mode on variables consider the following:
On selecting a variable, the number of dimensions of which does not match the number of output variable dimensions, the Change Dimension dialog box opens. Fix the dimensions that are missing in the output variable in this dialog box.
The following buttons for working with forecast correction variable are available:
Create. A variable (without data) is created that is used to correct the forecast. The variable frequency corresponds with the model frequency. The created variable is located in the root folder of modeling container. It is named Forecast Correction Factor and automatically opens for edit.
Fix. It fixes a variable. It opens the Change Dimension dialog box. If the forecast correction factor and output variable have matching dimensions, the button is not available.
Click the Calculate button to calculate a model by the specified equation. If during calculation any errors or warnings appear, they are displayed on the Errors and Warnings tabs respectively.
Examples of situations, at which a model can be calculated:
If the equation contains errors (invalid operators, non-existing element, and so on), an appropriate error message is displayed showing the number of the position containing this error. After the message closes, the cursor is placed to the indicated position of the equation.
If the equation can be calculated at least in one point.
See also:
Working with Input Variables | Time Series Analysis: Calculator