Consumer Fraud in Online Shopping: Detecting Risk Indicators through Data Mining

Consumer fraud in online shopping has become a major problem and severe challenge for online retailers. However, detection lags behind — for academia and practice — and data-driven knowledge about risk indicators in transaction data is still very limited. Thus, this study focuses on the empirical data-based identification of consumer fraud risk indicators and combinations in online shopping transaction data. We demonstrate the use of a decision tree as a data mining technique for analysis of data from one of the world's largest online retailers. Thereby, several patterns of fraud that improve separation of online shopping transactions into fraudulent and legitimate cases are identified. Thus, results can guide the choice of variables and design of fraud prevention actions and systems in future practical and theoretical work.