Support Vector Regression

Support Vector Regression (SVR) is a metamodeling technique prescribed for predictably high nonlinear behavior of the outputs with respect to the inputs.

SVR belongs to a general class of Support Vector Method (SVM) type techniques. These are data classification methods that use hyperplanes to separate data groups. The regression method works similarly. The main difference is that the hyperplane is used to categorize a subset of the input sample vectors that are deemed sufficient to represent the output in question. This subset is called the support vector set.

SVR uses non linear kernels.

Advanced Settings

You can access the SVR advanced settings in the MOP settings dialog box.

  1. Right-click the MOP node and select Edit from the context menu.

  2. Select the Use advanced settings check box.

  3. In the Advanced Settings tab, under Support Vector Regression, select the Use check box.

  4. Switch to the Support Vector Regression Settings tab.

  5. Click Change settings.

  6. Change the settings as required.

    NameDefault valueDescription
    Loss function typeEpsilon-insensitive

    Defines the loss function type. The options are:

    • Epsilon-insensitive: Ignores errors that are within epsilon distance of the observed value by treating them as equal to zero. The value of epsilon can affect the number of support vectors used to construct the regression function. With a bigger epsilon value, fewer support vectors are selected. On the other hand, a bigger epsilon value results in a more flat estimate.

    • Nu Epsilon-insensitive: Uses a parameter nu to control the number of support vectors.

    • Laplace: Absolute loss function.

    Epsilon0.02

    Specify the epsilon-tube within which no penalty is associated in the training loss function.

    Only available with Epsilon-insensitive loss function.

    The epsilon value must be between 1.e-6 and 0.999999.

    Nu0.5

    Controls the number of support vectors and training errors. Nu is an upper bound on the fraction of margin errors and a lower bound on the fraction of support vectors.

    Only available with Nu Epsilon-insensitive loss function.

    The epsilon value must be between 1.e-6 and 0.999999.

  7. To save and close the dialog, click OK.

Supported Versions

The following versions of SVR metamodel are supported and tested: 2021 R2 and later.