Create a Pattern Filter

You can add one pattern filter to a given transaction model. For that filter, you select a statistical function, define one or more sets of data the function applies to, and set parameters that focus the results.

Before you can create a pattern filter, you must select at least one business object for the model that has at least one attribute that provides data patterns can operate on. If not, then as you attempt to create the filter, an error message informs you that no patterns are associated with the business objects you've selected.

To create a pattern filter:

1. In the Model Logic panel, expand the Add Filter option, then select Pattern. A dialog box appears. Enter a name for the filter in the Name field.

2. Select one of the following patterns in the Pattern field:

   • Mean

   • Benford

   • Clustering

   • Anomaly Detection

   • Absolute Deviation

   • Pareto

   • Normalize

   • Lexical Tokenization

3. Under Model Objects, identify an attribute whose values are subject to analysis by the pattern. Select its business object in the Object field, and the attribute itself in the Attribute field.

   You can select more than one object-attribute pair: click the Add icon to generate additional selection fields. The fields display only objects and attributes your selected pattern can operate on.

4. Under Parameters, select parameter values. These vary by pattern.