Fraud Analytics: New Models, Real-Time Detection, and Sense-Making
Principal Investigator: Leman Akoglu, Assistant Professor of Information Systems, CMU Heinz College
PwC Sponsors: Orlando Lopez, Partner, Advanced Risk and Compliance Analytics, and Shane Foley, Partner, Advanced Risk and Compliance Analytics
Wherever there is money, there is fraud: healthcare fraud, auction fraud, money laundering, tax fraud, identity theft, credit card fraud, opinion fraud, and many other types. This project aims to tackle challenges associated with leveraging available data sources from multiple modalities in order to define different fraud models and formalize fraud-detection tasks. Besides designing new models for fraud detection, the project will also develop algorithms for each carefully defined fraud detection task, while addressing additional system challenges such as real-time detection, high throughput (i.e., fast processing time for incoming data), high accuracy (i.e., low error rates), and scalability. Lastly, the project will seek to produce visualizations and descriptions for the outcome of the designed algorithms to aid analysts in interpreting and taking action on the detected cases. Potential beneficiaries of developed technologies include government agencies (e.g., tax, social security), credit card companies, trading companies, healthcare companies, insurance companies.