Fraud Detection in Financial Services using Graph Analysis and Machine Learning by Hans Viehmann
Introduction at Spanish and Full session at English.
Both graph analysis and machine learning are effective techniques to detect anomalies and outliers in datasets. The former is particularly useful when relationships between entities play a role, in other words, when the data represents a network of connected things. Networks of bank accounts connected by financial transactions are one obvious example. For this reason, modern fraud prevention applications use graph analytics and pattern matching on this kind of data. In addition to this approach, financial service customers are increasingly complementing graph analysis with machine learning to yield even more accurate results. In this session, we will show how the Paysafe Group, a global provider of e-payment services, uses graph analysis for fraud detection. We will look at how to prepare relational data for graph analysis and how a machine learning platform can consume the linked data without losing the characteristics of the graph. Finally, we will explain why graph analysis and machine learning yield different results on the same data and discuss how data scientists can combine these techniques.