Basics (neural networks)

Neural network model. The type of model determines how the network connects the predictors to the targets through the hidden layer(s). The multilayer perceptron (MLP) allows for more complex relationships at the possible cost of increasing the training and scoring time. The radial basis function (RBF) may have lower training and scoring times, at the possible cost of reduced predictive power compared to the MLP.

Hidden Layers. The hidden layer(s) of a neural network contains unobservable units. The value of each hidden unit is some function of the predictors; the exact form of the function depends in part upon the network type. A multilayer perceptron can have one or two hidden layers; a radial basis function network can have one hidden layer.

• Automatically compute number of units. This option builds a network with one hidden layer and computes the "best" number of units in the hidden layer.
• Customize number of units. This option allows you to specify the number of units in each hidden layer. The first hidden layer must have at least one unit. Specifying 0 units for the second hidden layer builds a multilayer perceptron with a single hidden layer.
  Note: You should choose values so that the number of nodes does not exceed the number of continuous predictors plus the total number of categories across all categorical (flag, nominal, and ordinal) predictors.