Enterprise Miner^TM

An artificial neural network is a network of many simple processors ("units"), each possibly having a small amount of local memory. The units are connected by communication channels ("connections") that usually carry numeric (as opposed to symbolic) data encoded by various means. The units operate only on their local data and on the inputs they receive via the connections. The restriction to local operations is often relaxed during training.

More specifically, neural networks are a class of flexible, nonlinear regression models, discriminant models, and data reduction models that are interconnected in a nonlinear dynamic system. Neural networks are useful tools for interrogating increasing volumes of data and for learning from examples to find patterns in data. By detecting complex nonlinear relationships in data, neural networks can help make accurate predictions about real-world problems. The following neural network architectures are available in SAS Enterprise Miner:

• Generalized linear model (GLIM) -, which is suitable when there is a linear relationship between the target and the inputs.
• Multilayer perceptron (MLR) - default, which is often the best architecture for prediction problems
• Ordinary radial basis function* (RBF) with equal widths
• Ordinary RBF with equal widths
• Normalized RBF with equal heights
• Normalized RBF with equal volumes
• Normalized RBF with equal widths
• Normalized RBF with equal widths and heights
• Normalized RBF with unequal widths and heights

* Radial basis functions are often useful when the distribution between the targets and the inputs is normal.

Enterprise Miner's Neural Network GUI enables novice and advanced users alike to make appropriate modeling selections. For the novice user, most of the required network parameters are selected automatically based on the structure of the data and a few, easy settings. The advanced user, however, can drill down and change all network parameters using the GUI, which detects inadequate changes of interdependent parameters.

To avoid the tendency of neural networks to overfit the training data, the model performance is constantly assessed against the validation data, and the final model is selected based on one of several criteria that the user can select, such as the minimal validation error or maximum total profit. Using the interactive results browser, you can assess the performance of the models at every iteration and change the final model selection manually, if desired.

Output from the Neural Network node includes the following:

• Estimates data sets, for preliminary optimization and training.
• Output data sets, for training, validation, testing and scoring.
• Fit statistics data sets, for training, validation and testing.

Unlike other data mining solutions that limit you to a single algorithm, Enterprise Miner provides a full range of integrated models and algorithms. In addition to neural networks, Enterprise Miner provides decision trees, regression analyses, memory-based reasoning, bagging and boosting ensembles, two-stage models, clustering, time series, and associations.

Enterprise Miner combines a rich suite of integrated data mining tools with unprecedented ease of use, empowering users to explore and exploit corporate data for strategic business advantage.