Banking Fraud Solved by Graph Analysis
Most of us have seen the crime movies where the investigator has his/her cliche bulletin board with the lines of yarn linking various pieces of evidence. However, just because Hollywood likes to overuse the setup, doesn’t mean it’s effectiveness in the real world is devalued. Mapping crimes as graphs makes it much easier for investigators to see patterns within the data. Pieces of evidence become nodes, and common characteristics between the pieces of evidence become edges. It’s an intricate web of people, locations, dates, and other factors. Investigators need to know who talked to who, when and where events occurred, and so on. Its much harder to find relationships between pieces of evidence without the visual aid of a graph’s nodes and edges.
Unfortunately, many modern day crimes involve many people spread over multiple countries and can occur over the span of years. The old bulletin board in the office isn’t going to cut it. When fraudulent practices were discovered in the Swiss banking system, the data that investigators needed to sift through included over 100,000 clients in many different countries. Traditionally, when presented with large amounts of data, teams of investigators would manually comb through Excel files to try and find relationships. Luckily, Mar Cabra, an investigative journalist with experience in graph modeling, was part of the team designated to the Swiss banking frauds. He was able to lead a team to turn the data into a graph with over 275,000 nodes and 400,000 edges. Thanks to Cabra, the team created the bulletin board and yarn network at a much larger scale. With the innate visual properties of a graph, the investigative journalists were able to uncover frauds totaling a value in US dollars of over 100 billion.
One interesting thing to note is that the investigators purposefully avoided hiring data scientists. They believed that they could solve the cases without the need to pay for extra help. This means that they weren’t completely taking advantage of a graph’s potential. Graphs have many computational and mathematical techniques attributed to their structure, but this specific team focused only on using a graph as a visual tool.
http://www.computerworld.com.au/article/582489/fighting-fraud-graph-analytics/