An Attributed Graph Mining Approach to Detect Transfer Pricing Fraud
Alexey Tselykh, Margarita Knyazeva, Elena G. Popkova, Antonina V. Durfee, Alexander Tselykh · 2016
This paper presents an attributed graph-based approach to an intricate data mining problem of revealing affiliated, interdependent entities that might be at risk of being tempted into fraudulent transfer pricing. We formalize the notions of controlled transactions and interdependent parties in terms of graph theory. We investigate the use of clustering and rule induction techniques to identify candidate groups (hot spots) of suspect entities. Further, we find entities that require special attention with respect to transfer pricing audits using network analysis and visualization techniques in IBM i2 Analyst's Notebook.