Detecting Intentionality through Graph Mining

Tina Eliassi-Rad
Northeastern University
Computer Science & Network Science

We study a tax system (such as the one in Belgium) where companies have to pay employer and employee contributions to the government. Our study aims to identify those companies that intentionally declare bankruptcy in order to avoid contributing their taxes -- i.e., they commit fraud. We link companies to each other through their shared resources, as some resources are the instigators of fraud. We introduce GOTCHA!, a new approach on (1) how to define and extract features from a time-weighted network, and (2) how to exploit and integrate network-based and intrinsic features for fraud detection. The GOTCHA! propagation algorithm diffuses fraud through the network, labeling the unknown and anticipating future fraud whilst simultaneously decaying the importance of past fraud. We find that domain-driven network variables have a significant impact on detecting fraud, and improve on baselines by detecting up to 55% additional fraudsters over time. This is joint work with Véronique Van Vlasselaer (KU Leuven & SAS), Leman Akoglu (Stony Brook & CMU), Bart Baesens (KU Leuven), and Monique Snoeck (KU Leuven).


Back to Workshop III: Cultural Patterns: Multiscale Data-driven Models