Data mining — Working with mining models and test results

You can explore mining models and apply them to new data. You can also deploy mining models or test results by extracting information.

At the end of the model-building process, you must decide how to deploy the test results or the models that you have created. You might also need to convert a model to the data type that can be recognized by the application that you want to use.

Intelligent Miner® provides user-defined functions to do the following tasks:

Retrieving information from a mining model
Exporting your mining model to a CLOB value
Accessing the results of a classification test run or a regression test run
Exporting the test result of a classification test run or a regression test run to a CLOB value

The following list shows where you can deploy the model and how you can prepare the model content for the corresponding application phase. You can deploy the model in the following applications:

IM Visualization

To graphically view and analyze the results. You can view, for example, gains charts, confusion matrix, field importance, and quality views.

Intelligent Miner

To compute scores for new records.

To prepare your model for scoring, you can retrieve information from the model, for example the signature of a model and the importance of its field.

The following table shows the data types you can retrieve information from and the sections that list the appropriate functions to be used:

Table 1. Data types you can retrieve information from
Data type	Section
`DM_ClasModel`	Functions for working with model type DM_ClasModel
`DM_ClasTestResult`	Functions for working with test results
`DM_ClusteringModel`	Functions for working with model type DM_ClusteringModel
`DM_RegressionModel`	Functions for working with model type DM_RegressionModel
`DM_RegTestResult`	Functions for working with test results
`DM_RuleModel`	Functions for working with model type DM_RuleModel
`DM_TimeseriesModel`	Functions for working with model type DM_TimeseriesModel

Other applications

Any other application that can retrieve the model contents from tables that are created by user-defined functions in Intelligent Miner. These tables can provide, for example, the following information to be processed:

Clusters and the description of a Clustering model.
Rules as database rows, including the bodies and names of rules for an Associations model.
One record for each node of a tree classification model. Each record contains the node ID, as presented by the visualizer, a textual description of the rule that defines the node. In addition to this, the predicted class, the confidence, and the depth of the node are provided.
Correlations between fields of a data mining model. The correlation of some pairs of fields is shown, such as the pair of fields for which a correlation has been calculated when building the model.
The signature of a model (that means the types and names of all data fields which are used in the model) and the importance of its fields for the resulting model.