You must define mining settings for your training mining
run. The mining settings contain the specific instructions for the
mining function that you want to use. The definitions and roles of
the fields also belong to the settings specification. This means that
the logical data specification is part of the mining settings.
The following list shows the mining functions that provide mining-specific
settings:
- Clustering
- You can specify the maximum number of clusters to be created or
the similarity threshold for a distribution-based clustering training
run.
- Associations
- You can reduce to your preferred level the number of associations
that are created.
- Classification and Regression
- You must specify the name of the target field.
Defining mining settings in the model-building process means to
create the following data types and to perform operations on these
data types:
- DM_RuleSettings
- DM_ClasSettings
- DM_ClusSettings
- DM_RegSettings
- DM_TsSettings
The data types store the settings for associations mining runs,
classification mining runs, clustering mining runs, regression training
runs, or sequence rules mining runs. These mining settings are sets
of control parameters.
You can limit the fields that are used for mining by manually indicating
the fields to be used.
You can also use automatic preprocessing and customization to identify
the fields to be used or to be dropped. Using the automatic preprocessing
includes the following advantages:
- Reducing the mining skills that are necessary to deploy mining
functions
- Invoking all algorithms on arbitrary input tables without manual
preprocessing and elaborate settings customization
- Providing default settings to ensure good run-time results and
models