SVM Node Expert Options

If you have detailed knowledge of support vector machines, expert options allow you to fine-tune the training process. To access the expert options, set Mode to Expert on the Expert tab.

Append all probabilities (valid only for categorical targets). If selected (checked), specifies that probabilities for each possible value of a nominal or flag target field are displayed for each record processed by the node. If this option is not selected, the probability of only the predicted value is displayed for nominal or flag target fields. The setting of this check box determines the default state of the corresponding check box on the model nugget display.

Stopping criteria. Determines when to stop the optimization algorithm. Values range from 1.0E–1 to 1.0E–6; default is 1.0E–3. Reducing the value results in a more accurate model, but the model will take longer to train.

Regularization parameter (C). Controls the trade-off between maximizing the margin and minimizing the training error term. Value should normally be between 1 and 10 inclusive; default is 10. Increasing the value improves the classification accuracy (or reduces the regression error) for the training data, but this can also lead to overfitting.

Regression precision (epsilon). Used only if the measurement level of the target field is Continuous. Causes errors to be accepted provided that they are less than the value specified here. Increasing the value may result in faster modeling, but at the expense of accuracy.

Kernel type. Determines the type of kernel function used for the transformation. Different kernel types cause the separator to be calculated in different ways, so it is advisable to experiment with the various options. Default is RBF (Radial Basis Function).

RBF gamma. Enabled only if the kernel type is set to RBF. Value should normally be between 3/k and 6/k, where k is the number of input fields. For example, if there are 12 input fields, values between 0.25 and 0.5 would be worth trying. Increasing the value improves the classification accuracy (or reduces the regression error) for the training data, but this can also lead to overfitting.

Gamma. Enabled only if the kernel type is set to Polynomial or Sigmoid. Increasing the value improves the classification accuracy (or reduces the regression error) for the training data, but this can also lead to overfitting.

Bias. Enabled only if the kernel type is set to Polynomial or Sigmoid. Sets the coef0 value in the kernel function. The default value 0 is suitable in most cases.

Degree. Enabled only if Kernel type is set to Polynomial. Controls the complexity (dimension) of the mapping space. Normally you would not use a value greater than 10.