Previous Page | Next Page

Working with Nodes That Modify, Model, and Explore

Develop Other Competitor Models


Overview

Enterprise Miner enables you to try several different models and then compare the results in one common framework. You can then choose the best model for scoring new data.

One method of arriving at a model is the artificial neural network. The artificial neural network attempts to mimic the actions of the human brain, which is a large organic neural network. Experienced data miners know that artificial neural networks, when carefully tuned, are often very useful in showing nonlinear associations between inputs and the target. Common applications for neural network models include credit risk assessment, direct marketing, and sales predictions.


Add a Neural Network

In this task, you use the Neural Network node to build a neural network model that is based on your data.

  1. Drag a Neural Network node from the Model tab of the node toolbar into the Diagram Workspace, and connect it to the Transform Variables node.

    [untitled graphic]

  2. Select the Neural Network node in the Diagram Workspace; then set the following Neural Network properties in the Properties panel:

    • Click the ellipsis icon in the Network property.

      [untitled graphic]

      This opens the Network window.

      [untitled graphic]

    • Set the Direct Connection property to Yes.

    • Set the Number of Hidden Units property to 5.

  3. Run the Neural Network node.


Add an AutoNeural Model

The AutoNeural node enables you to find optimal configurations for a neural network model.

  1. Drag an AutoNeural node from the Model tab of the node toolbar into the Diagram Workspace, and connect it to the Variable Selection node. The combinatorial search that the AutoNeural node performs can be computationally expensive, so you should reduce the input set for training.

    [untitled graphic]

  2. Select the AutoNeural node in the Diagram Workspace, and then set the following properties in the AutoNeural Properties panel:

    • Set the Architecture property to Cascade, in order to add nodes in a cascade fashion.

    • Set the Train Action property to Search, in order to add nodes according to the Cascade architecture and to find the best topology for the network.

    [untitled graphic]

  3. Run the AutoNeural node.

Previous Page | Next Page | Top of Page