If you haven’t already – please refer to the previous post on an introduction to Social Network Analysis to help provide basics to understanding what is discussed in this post.
Just like how SAS is used to analyse text analytics, there are also a variety of software used to analyse Social Network Analysis (SNA). Some run in the cloud e.g. Polinode, SocioViz, and others run locally e.g. NodeXL, R.
We will take a closer look at exactly what you can learn from SNA. One area that SNA has been put into use in fraud detection. As fraud tends to follow similar patterns, the network of actors in fraud is likely to follow the below variations:
First off, the density of a network can be analysed for any signs of fraud. Density measures can take any value between 0 to 1, with 0 being the lowest density, and 1 the maximum. Fraud hotspots will register a higher density due to increased activity from account transactions, money laundering, and transaction monitoring. Nodes can then be flagged, and the network of links related to only these nodes are generated. Clusters are built using the snowball method until no further nodes are identified. Visual patterns and links are observed to generate measures for network features, and finally if required, mitigation measures can be applied.
In an example of possible fraud in a car dealership, the steps above would be followed out as below:
- Suspicious dealers (nodes) that were flagged would have their links expanded using the snowball method to help identify cars and bank accounts involved
- The fraudulent relationships between dealers were identified as clusters, and density measures applied to see where the areas with the highest fraudulent activity was
- Centrality measures (i.e. degree, closeness, betweeness, and eigenvector) were applied to the identified clusters to identify the key actors in each cluster
- Banks then took the measure of confronting the involved dealers, breaking up the fraud clusters, prevent future financing of vehicles that fall into the above suspicious patterns, and also ensuring that only the obvious actors causing the fraud were penalised, and not legitimate dealers
Example of how the fraudulent dealers were identified
Social Network Analytics can be used in conjunction with Social Media Analytics to help businesses gain insight about their customers. How do you think that the use of both strategies can help businesses as opposed to using one or the other? Let us know in the comments below!
- Cambridge Intelligence. (2016). Social Network Visualization – Cambridge Intelligence. [online] Available at: http://cambridge-intelligence.com/keylines/social-network-analytics/ [Accessed 8 May 2016].
- Freeman, L. C., 1978. Centrality in social networks: Conceptual clarification. Social Networks 1, 215-239
- Kirchner, C. and Gade, J. (2011). Implementing social network analysis for fraud prevention. 1st ed. [ebook] Düsseldorf: CGI. Available at: https://www.cgi.com/sites/default/files/white-papers/Implementing-social-network-analysis-for-fraud-prevention.pdf [Accessed 8 May 2016].
- Krebs, V. (2015). Social Network Analysis: An Introduction by Orgnet,LLC. [online] Orgnet.com. Available at: http://www.orgnet.com/sna.html [Accessed 8 May 2016].