Fraud Analytics Using Descriptive, Predictive, and Social Network Techniques: A Guide to Data Science for Fraud Detection

Chapter 6Fraud Analytics: Post-Processing

Introduction

The result of the analytics step is an analytical fraud model built using either a descriptive, predictive, social network or combined technique. Essentially, the analytical model reduces to a mathematical formula predicting the fraud occurrence or fraud amount. In a next step, this model or formula needs to be integrated into the existing business environment and ICT architecture. In order to successfully complete this exercise, it is of key importance to perfectly understand the business requirements, which are usually specified by the end users of the analytical model(s). Furthermore, after the models have been put to work, they need to be closely monitored such that any deviation in performance due to changing fraud behavior can be detected in a timely way and corresponding actions can be undertaken. In this chapter, we will discuss various issues that arise when deploying, using, and monitoring analytical fraud models within a specific business context.

The Analytical Fraud Model Life Cycle

In previous chapters, we discussed how to develop analytical fraud models using descriptive, predictive, and social network analytics. Once the model development has been completed, the analytical model can proceed to the next step in the model life cycle (see Figure 6.1).

Figure 6.1 The Analytical Model Life Cycle

During model ...