Sampling Implementation

Specify sampling settings to determine sample sizes or rely on algorithm-calculated sample sizes for efficient rule generation.

Usage Notes

1. If ODMS_SAMPLING is unspecified or set as ODMS_SAMPLING_DISABLE, the sampling is not performed for AR and the exact AR is obtained.

2. If ODMS_SAMPLING is set as ODMS_SAMPLING_ENABLE and if ODMS_SAMPLE_SIZE is specified as positive integer number then the user-specified sample size (ODMS_SAMPLE_SIZE) is utilized. The sampling is performed in the general data preparation stage before the AR algorithm. The AR-specific sample size is not computed. The approximated AR is obtained.

3. If ODMS_SAMPLING is set as ODMS_SAMPLING_ENABLE and ODMS_SAMPLE_SIZE is not specified, the AR-specified sample size is computed and then sampling is performed in the AR algorithm. The approximated AR is obtained.

   Note:

   If the computed AR-specific sample size is larger than or equal to the total transaction size in the data set, the sampling is not performed and the exact AR is obtained.

If users do not have a good idea on the choice of sample size for AR, it is suggested to leave ODMS_SAMPLE_SIZE unspecified, only specify proper values for sampling parameters and let AR algorithm compute the suitable AR-specific sample size.

See Also:

DBMS_DATA_MINING — Machine Learning Function Settings for a listing and explanation of the available model settings.

Note:

The term hyperparameter is also interchangeably used for model setting.