Using Social Network Analysis for Fraud Detection

Abstract

Facing the age of digitalisation, inspectorates are changing their way of working using a risk-oriented and data-driven approach. Social Network Analysis (SNA) seems to be promising in detecting fraud and simultaneously contributing to increasing the effectiveness and efficiency of inspection work. SNA emphasises the structural aspects of networks to detect and interpret patterns of social entities, both graphically and mathematically. Although the emergence of big data opens great opportunities in the public domain and the benefits of using SNA in fraud detection seem to be clear cut, a more institutional view challenges the assumption that simply working data-driven leads to better deployment of enforcement assets. In practice, the way in which business value is created from big data often remains unclear. There seems to be a gap between the promises of big data and its practical realisations, in particular in the public domain. Therefore, this research considered big data in the context of SNA from an institutional perspective. This means it is assumed that actors that shape the process from data as a raw material to the final deployment of inspection capacity based on the outcome of the analysis. This research underlined a decision-making perspective which states that the way in which the alternatives are framed impacts the alternative chosen by people and in turn the subsequent decision. By using a qualitative research approach consisting of a multiple-case study design combined with action research this research contributed to two main purposes. First of all, the study had a functional purpose aimed to explore the use of SNA in fraud detection, more specifically fraud in the context of food and consumer products. Following from the analysis of two large real-world data sets, it turned out that network visualisation offers a powerful solution to make information hidden in networks easy to interpret and understand. With one glance at the network one could identify who does business with whom, which entities act as bridges between two clusters, trace suspicious patterns, and gain insight into the overall structure of the networks. Applying network metrics helped to quickly identify the important players in the networks and could be used to evaluate or predict the possible consequences of removing specific actors from the networks to destabilise the networks. Secondly, the research had an institutional purpose aimed to get insight into big data value creation in the public sector. During the research, it became evident that important assumptions and decisions have been which appeared to be fundamental for the outcome of the analysis. This prevented the creation of options or led to options that were sub-optimal. Neglecting them would be at the detriment of any SNA-ambition an organisation may hold. As the research into both areas is still in a very young state, the findings from this thesis form a starting block for other studies to expand on. Future research will be addressed to widen the empirical evidence on how big data affects public decision-making.