Using Model Nuggets in Streams

Model nuggets are placed in streams to enable you to score new data and generate new nodes. Scoring data enables you to use the information gained from model building to create predictions for new records. To see the results of scoring, you need to attach a terminal node (that is, a processing or output node) to the nugget and execute the terminal node.

For some models, model nuggets can also give you additional information about the quality of the prediction, such as confidence values or distances from cluster centers. Generating new nodes enables you to easily create new nodes based on the structure of the generated model. For example, most models that perform input field selection enable you to generate Filter nodes that will pass only input fields that the model identified as important.

To Use a Model Nugget for Scoring Data

1. Connect the model nugget to a data source or stream that will pass data to it.
2. Add or connect one or more processing or output nodes (such as a Table or Analysis node) to the model nugget.
3. Execute one of the nodes downstream from the model nugget.

Note: You cannot use the Unrefined Rule node for scoring data. To score data based on an association rule model, use the Unrefined Rule node to generate a Rule Set nugget, and use the Rule Set nugget for scoring. See the topic Generating a Rule Set from an Association Model Nugget for more information.

To Use a Model Nugget for Generating Processing Nodes

1. On the palette, browse the model, or, on the stream canvas, edit the model.
2. Select the desired node type from the Generate menu of the model nugget browser window. The options available will vary, depending on the type of model nugget. See the specific model nugget type for details about what you can generate from a particular model.