Change the Discretization of a Column in a Mining Model
Topic Status: Some information in this topic is preview and subject to change in future releases. Preview information describes new features or changes to existing features in Microsoft SQL Server 2016 Community Technology Preview 2 (CTP2).
Analysis Services automatically discretizes values—that is to say, it bins data in numeric column—in certain scenarios. For example, if your data contains continuous numeric data and you create a decision tree model, each column of continuous data will be automatically binned, depending on the distribution of the data. If you want to control how the data is discretized, you must change the properties on the mining structure column that control how the data is used in the model.
For general information about how to set the properties in a mining model, see Mining Model Columns.
To display the properties for a mining model column
In the Mining Models tab in Data Mining Designer, right-click the column header that contains the name of the mining model, or the row in the grid that contains the name of the mining algorithm, and then select Properties.
The Properties window displays the properties that are associated with the mining model as a whole.
In the Structure column near the left side of the screen, click the column that contains the continuous numeric data you want to discretize.
The Properties window changes to display just the properties associated with that column.
To change the discretization method
In the Mining Properties window, click the text box next to Content, and select Discretized from the dropdown list.
In the Properties window, click the text box next to DiscretizationMethod and select one of the following values: Automatic, EqualAreas, or Cluster.
If the column usage is set to Ignore, the Properties window for the column is blank.
The new value will take effect when you select a different element in the designer.
In the Properties window, click the text box next to DiscretizationBucketCount and type a numeric value.
If you change these properties, the structure must be reprocessed, along with any models that you want to use the new setting.