Visual Fraud Analysis, supporting visual communication between data and analyst

Abstract

In the current information age the Internet allows us to gather and analyze more data than ever which can easily lead to a persons’ cognitive limits. With data visualization one could relieve cognitive load, which means that the analysts will have more cognitive power to think about patterns or relations within the dataset. In this study, Design-Based Research methodology helped to iteratively develop both a theoretical and a practical result. Firstly, a visualization model build from theory for designing a visual support tool for analysts in general and secondly, a proof of concept of a visual support tool for a specific group of analysts (fraud analysts). The visualization model encompasses the scope from low-level perceptual processes to high- level cognitive processes and explains these from different perspectives (i.e. the analysts, their interaction with the visual tool, design elements and the visual communication designer). The model offers guidelines on how other visual support tools could be designed and therefore offers procedural power. Additionally the model offers explanations why such a design could aid visual analysis and thereby shows descriptive power. The proof of concept allows analysts to not only confirm or reject their hypotheses on frauds but it also gives the options for exploring alternative hypotheses. This is done by visually showing different levels within the data (i.e. overview on variables, specific relation between variables, distribution of one variable) and by offering useful and fluent interaction.