Oracle Minimum Description Length (MDL)

The Oracle Minimum Description Length (MDL) algorithm helps to identify the attributes that have the greatest influence on a target attribute. Oftentimes, knowing which attributes are most influential helps you to better understand and manage your business and can help simplify modeling activities. Additionally, these attributes can indicate the types of data you may wish to add to augment your models. MDL might be used, for example, to find the process attributes most relevant to predicting the quality of a manufactured part, the factors associated with churn, or the genes most likely involved in the treatment of a particular disease.

Oracle MDL discards input fields that it regards as unimportant in predicting the target. With the remaining input fields it then builds an unrefined model nugget that is associated with an Oracle model, visible in Oracle Data Miner. Browsing the model in Oracle Data Miner displays a chart showing the remaining input fields, ranked in order of their significance in predicting the target.

Negative ranking indicates noise. Input fields ranked at zero or less do not contribute to the prediction and should probably be removed from the data.

To display the chart

  1. Right-click on the unrefined model nugget in the Models palette and choose Browse.
  2. From the model window, click the button to launch Oracle Data Miner.
  3. Connect to Oracle Data Miner. See the topic Oracle Data Miner for more information.
  4. In the Oracle Data Miner navigator panel, expand Models, then Attribute Importance.
  5. Select the relevant Oracle model (it will have the same name as the target field you specified in IBM® SPSS® Modeler). If you are not sure which is the correct one, select the Attribute Importance folder and look for a model by creation date.