Interactive Rule Refinement for Fraud Detection

Credit card frauds are unauthorized transactions that are made or attempted by a person or an organization that is not authorized by the card holders. Fraud with general-purpose cards (credit, debit cards etc.) is a billion dollar industry and companies are therefore investing significant efforts in identifying and preventing them. It is typical to deploy mining and machine learning-based techniques to derive rules. However, such rules may not always capture the semantic reasons underlying the frauds that occur. For this reason, credit card companies often employ domain experts to manually specify rules that exploit general or domain knowledge for improving the detection process. Over time, however, as new (fraudulent and legitimate) transactions arrive, these rules need to be updated and refined to capture the evolving (fraud and legitimate) activity patterns. The goal of the RUDOLF system described in this paper is to guide and assist domain experts in this challenging task. RUDOLF automatically determines the “best” adaptation to existing rules to capture all fraudulent transactions and, respectively, omit all legitimate transactions. The proposed modifications can then be further refined by users and the process can be repeated until they are satisfied with the resulting rules. We show that the problem of identifying the best candidate adaptation is NP-hard in general and present PTIME heuristic algorithms for determining the set of rules to adapt. We have implemented our algorithms in RUDOLF and show, through experiments on real-life datasets, the effectiveness and efficiency of our solution.