NEVA: Visual Analytics to Identify Fraudulent Networks

Abstract

Trust‐ability, reputation, security and quality are the main concerns for public and private financial institutions. To detect fraudulent behaviour, several techniques are applied pursuing different goals. For well‐defined problems, analytical methods are applicable to examine the history of customer transactions. However, fraudulent behaviour is constantly changing, which results in ill‐defined problems. Furthermore, analysing the behaviour of individual customers is not sufficient to detect more complex structures such as networks of fraudulent actors. We propose NEVA (Network dEtection with Visual Analytics), a Visual Analytics exploration environment to support the analysis of customer networks in order to reduce false‐negative and false‐positive alarms of frauds. Multiple coordinated views allow for exploring complex relations and dependencies of the data. A guidance‐enriched component for network pattern generation, detection and filtering support exploring and analysing the relationships of nodes on different levels of complexity. In six expert interviews, we illustrate the applicability and usability of NEVA.

A Survey of Human‐Centered Evaluations in Human‐Centered Machine Learning

“…While papers describe interactions at varying levels of detail, a few common interaction types emerged in our analysis, including selection (e.g., [KBJ∗20; WBL∗20]), filtering (e.g., [LGM∗20; PNKC20]), zooming (e.g., [LPH∗20; LSC∗18]), tuning weights/hyperparameters (e.g., [DSKE20; SLC∗20]), and annotation/labeling (e.g., [SSKE19; XXM∗19]). Multiple of these interaction types tend to be combined in a single system, for example when users first select a set of data instances before labeling them (e.g., [BHZ∗18; ZWLC19]).…”

“…Fraud Detection – In cooperation with a bank, Leite et al [LGM∗20] develop an environment to explore fraudulent behavior and state that “ trustability, reputation, security and quality are the main concerns for public and private financial institutions ” [LGM∗20]. They derive four design requirements with collaborating experts and evaluate the success of the system using four tasks, each targeting one requirement.…”

“… Clearly state the research questions that you are aiming to investigate (e.g., [BHZ∗18; HHC∗19; LGM∗20]). Structure the evaluation section to support readability, e.g., [HHC∗19; SFB∗20]. Highlight key findings. If applicable, represent a study's phases as a figure, such as provided by Ren et al [RAL∗17] (depicted in Figure 4).…”

“…69% of papers that mentioned guidance in their evaluation described systems with orienting guidance (e.g., [ESD∗19; SSKE19; WMJ∗19]) and 18% and 9% report insights into directing [KEV∗18; LGM∗20; SMD∗16] and prescribing guidance [EKC∗20; GLC∗19], respectively. As guidance is not typically evaluated in‐depth, we observed no difference between the qualitative feedback for different guidance degrees.…”

Guide Me in Analysis: A Framework for Guidance Designers

“…Overview visualization for finding frauds in a money transactions network [LGM19]. The analytical exploration of the network data is supported by allowing analysts to formulate only semantically relevant queries and make the analysis effective.…”