Graph analysis and machine learning provide different approaches to anomaly detection and fraud prevention and are now included with the Oracle Database free of charge. In this talk, we’ll look at two customer cases and will discuss decision criteria on when to use which technology.
Both graph analysis and machine learning can be used very effectively to detect anomalies and outliers in datasets. The former is particularly useful when data can be represented as a network in which the connectedness of data, ie. the explicit relationships between entities, play a role. Networks of bank accounts connected by financial transactions are one obvious example, which is why modern fraud prevention applications use graph analytics and pattern matching on this kind of data. Graph analysis can be complemented by machine learning to yield results that are even more accurate. In this session, we will show how the Paysafe Group, a global provider of e-payment services, is using graph analysis for the purpose of fraud detection. We will look at how graph analysis and machine learning yield different results and how these techniques can be combined, provided the linked data can be turned into data structures that can be used in machine learning while at the same time maintaining the characteristics of the graph.