Creating an Approximation Using the RSM Model

You can create an approximation using the Response Surface Method model and tune the approximation options. You can use the Predict page to compare the quality of your approximations.

See Also
Approximations
The Response Surface Approximation Model
Understanding Your Approximation
  1. From the Approximations table on the left side of the Predict page, select .

    Note: If no approximations have been created, you can also click Create New Approximation.

    The Create New Approximation dialog box appears.
  2. From the Algorithm tab, enter a name for the approximation.
  3. Select Response Surface Method from the Approximation Type options,

    Note: If desired, you can create approximations using both models and compare the approximations side by side or in an overlay plot to determine which model produced the best numerical solution.

  4. Enter a value for the F-ratio to Add Term, the minimum value of the F-ratio to add a new polynomial term to the approximation model ( F e n t e r ).
  5. Enter a value for the F-ratio to Drop Term, the maximum value of the F-ratio to drop a polynomial term from the approximation model ( F d e l e t e ).
  6. Enter the maximum number of data points used during the calculation of the approximation.

    By default, Results Analytics calculates the approximation using a maximum number of 2000 data points. The accuracy of the approximation increases with more data points; however, the length of time required for Results Analytics to fit the approximation also increases. You might have to experiment with the number of points to find a balance between accuracy and speed.

    If your data set contains more data points than the maximum specified, Results Analytics searches through the entire data set and selects points randomly to include in the approximation.

  7. Enter a value for the Polynomial Order of the Response Surface Method model (1 ≤ value ≤ 10).

    You can enter a value of 1 if you know that there is a linear relationship between the inputs and the outputs of the approximation. A value of 2 is recommended for most cases and provides the best performance to cost ratio as well as the best optimization performance for smooth exact functions. Alternatively, you can enter the maximum value of 10 and allow Results Analytics to eliminate unnecessary terms.

    Check the measure of fit information for signs the approximation is overfitting the data—the cross validation percent errors are much higher than the residual percent errors. If the approximation is overfitting, you should recalculate the approximation with lower-order polynomials.

  8. From the Validation tab, set the validation type and validation hyperparameters such as Number of Folds for K-Folds Cross Validation.
  9. From the Inputs tab, select input parameters to exclude from the generation of a specific approximation.

    These inputs will not be used in training and will not be displayed in the prediction profiler.

  10. From the Datasets tab, select datasets to exclude from approximation generation.
  11. Click Create to create the approximation.
    The approximation is created with the name you specified and appears in the list of approximations.