1. Using Advanced Controls
  2. Interpret Transaction Model Results

Interpret Transaction Model Results

The results returned by a transaction model depend on whether that model includes a pattern filter.

Defined Models

If the model doesn't include a pattern filter, it returns a grid. Each row is the record of a transaction that violates the model, although a single violation may encompass multiple rows. For example, a model that detects duplicate invoices would return one row for each of the duplicated invoices.

Each record contains values for the result attributes selected for the model, and for any derived attributes the model may calculate. The grid may also contain these columns:

  • Incident Information. For each record, this is the value of the first attribute among those selected to characterize the suspect transaction.
  • Group and Grouping Value. Values for these columns vary:
    • A filter may use the Equals condition to set an attribute of a business object equal to itself. The Group filter reports the business object and attribute. The Grouping Value field reports the value of the attribute.
    • A filter may find transactions with similar values for a specified attribute. The Group field displays the word "Similar" and the specified attribute. The Grouping Value field displays the value of the attribute.
    • A function filter may calculate a value for a specified attribute across a group of transactions. The Group field identifies the function and the specified attribute. The Grouping Value field displays the calculated value.

Pattern Models

If the model contains a pattern filter, it typically generates both graphic and tabular results. The graph depicts the statistical pattern generated by the model, and the table displays data represented in the graph. (The Normalize and Lexical Tokenization patterns are exceptions. Each generates only tabular results.)

For example, the graph for a model that uses the Mean pattern displays two plots: One represents the mean of the values of an attribute. The other tracks means for specified subsets of those values, with outliers evident by their distance from the overall mean.

If a pattern analysis uses multiple attributes, the results page generates multiple result tabs. Each presents a graph and a table related to one of the attributes. If you hold the mouse cursor over a data point in a graph, a box displays the values defining that point. If you click on the data point, the table refreshes to display only values for the data point you've selected.