1. Using Advanced Controls
  2. Overview of Transaction Models

Overview of Transaction Models

A transaction model uncovers transactions that might involve error or fraud, or otherwise present risk.

It consists of filters that define aspects of risk and select records that satisfy their definitions. A combination of these filters defines a complete risk, with each filter evaluating records returned by filters that precede it.

Filters cite business objects and attributes of those objects, which supply data for analysis. Oracle provides "delivered" business objects. Each is a set of related fields from a business application, and an attribute is one field within the set. You can create other types of object: An imported object is data imported via an xml file. A user-defined object is data returned by a specially configured advanced control. A system-generated object is data returned by certain transaction filters.

Standard Filters

A standard filter selects records containing an attribute whose values satisfy a condition. For example, a standard filter may state: Payment Amount (an attribute of the Payment business object) is greater than 5,000 dollars. As you create the filter, you:

  • Specify the attribute (and the object it belongs to).

  • Select the condition from a set of predefined conditions. (In the example, "is greater than" is the condition.)

  • Specify elements that complete the condition. These elements may be one or more constant values or another attribute of an object. (The example uses a single constant value, "5,000 dollars.")

Certain conditions enable a standard filter to gather attribute values into groups. For example, a filter may use a Similar condition to find invoices with similar supplier names. It would return sets of records, each set containing invoices that meet a similarity standard in a distinct way. The filter in this example may serve to identify duplicate invoices with slightly different renderings of a supplier's name.

Function Filters

A function filter, like a standard filter, creates a formula that specifies how attribute values must satisfy a condition. However, it also incorporates a function that operates on the attribute term, for example taking the average of a set of attribute values. It can create groups of records for the function to operate on, or it can use groups created by a standard filter.

For example, a function filter might group records by supplier. It may then calculate an average payment amount for each supplier, then determine whether each average amount exceeds a threshold value.

Pattern Filters

A pattern filter performs statistical analysis. As you create the filter, you select a pattern (a statistical function) from a predefined set, and you select one or more attributes whose values are subject to analysis. For example, a Mean pattern calculates the average for a set of numeric attribute values, and identifies values too far above or below the average. Each model can use only one pattern filter.