Download (Simple) Fraud Detection Model

The figure below illustrates the simple fraud detection model used on this web site.

The nodes of the graph specify variables, the variables on interest here being "Fraud". The other variables represent pieces of information about a customer that may or may not be known. The edges (arrows) in the graph represent dependence relations between pairs of variables.

Attached to each node in the model is a conditional probability distribution which defines the strengths of the dependence relations between a variable and its parents in the graph.

This very simplified model merely serves to describe the principles behind a HUGIN fraud model. A model used in a production environment will typically consist of 20 to 40 variables.

Please download the fraud prediction model as a HUGIN Network Language Specification file.

Download Example from Technical White Paper

The figure below illustrates the fraud dection model used in the HUGIN Fraud Detection Management Technical White Paper.

Please download the fraud prediction model as a HUGIN Network Language Specification file and the Java code example

To run the example you need access to the HUGIN Java API. Alternatively, you can submit this form and download a zip-archive containing the example model, the Java code example, and the HUGIN Lite Java API libraries for the Microsoft Windows platforms.