Select multilayer perceptron (MLP) or radial basis function (RBF). Both use feedforward architectures—data only moves from input nodes through the hidden layer of nodes to output nodes.
Information about the neural network is displayed visually, including the dependent variables, number of input and output units, number of hidden layers and units and activation functions.
Choose to display results in tables or graphs. Save optional temporary variables to the active dataset. Export models in XML-file formats to score future data.
Control the process
Specify the dependent variables, which may be scale, categorical or a combination of the two. Adjust each procedure by choosing how to partition the data set, which architecture to use and what computation resources to apply to the analysis.
Combine with other procedures
Confirm neural network results with traditional statistical techniques. Gain clearer insight in a number of areas, including market research, database marketing, financial analysis, operational analysis and health care.