You can refine the way that the target value is handled in several ways; not all of these ways are appropriate for all algorithms:
When you run the wizard from a build activity, you specified some of this information when you built the model. The activity creates weights or a priors table, as appropriate.
You can select whether a target value should be weighted during scoring (model apply). The default is to select no such weights value.
If you selected maximum average accuracy for a decision tree or support vector machine model, this page allows you to view or modify the cost matrix settings (decision tree) or the weights (support vector machine).
To view or change costs, click Cost Matrix Settings.
To view or change weights for an SVM model, click Weights.
To specify the Weight, either accept the default or specify a number less than or equal to 20.0.
The wizard generates a cost matrix that will be used when the model is applied to new data. For more information about costs, see Costs.
For more information about weights and costs, see Oracle Data Mining Concepts in Where to Find More Information.
For Naive Bayes models, you can specify how the sample differs from the population by providing priors. The default is to provide no priors.
To specify priors, click Priors Settings and supply the information.
If you run this wizard from an activity, the activity has created a priors table. You can modify this priors table.
Click Next to continue. The wizard performs the following validations:
Copyright © 2006, 2008, Oracle. All rights reserved.