Entry Date:
September 25, 2015

Situational Awareness Framework for Ranking-Based Financial Fraud Detection (SAFARI)

Principal Investigator Abel Sanchez


Situational Awareness framework that exploits different perspectives of the same financial data and assigns risk scores to entities (e.g. payment documents) to improve false positive ratios and assist the identification of fraudulent activity in huge and unlabeled financial data.

A novel Red Flag Network (RFNet) link analysis approach allows SAFARI to integrate the data generated by a variety of anomaly detectors. An RFNet is a network of entities (e.g. payment documents) where entities are connected if they participate in automatically raised red flags.

The SAFARI software platform can be upgraded and extended to take advantage of new advances and add more capabilities in data and visual analytics for fraud detection. Being a flexible software framework, SAFARI can keep up with business evolution and dynamics, having the potential to be a valuable advisor in decision-making in both the short and the long-term.

SAFARI's treemaps, RFNet and map data visualizations are enablers that provide insights into the related entities being inspected and unveil relationships that are hard to be discovered otherwise. The objective is to help SMEs in perceiving patterns, gain insights into the data and make sense of disparate but potentially related phenomena.

Fraud schemes arise when the red flags in RFNets are related and the case is solved when the loose ends are tied up. To relate red flags, SAFARI implements a Bayesian Belief Network (BBN) apporach for advanced evidence fusion and risk ranking. A BBN is a probabilistic graphical model that allows SAFARI to deal with fraud occurrence uncertainly by representing causal relationships among financial red flags to which observed occurrence is posted as evidence.