Join us February 4, 2014 at 9am PDT for our latest DSC's Webinar Series: Machine Learning For Forensic Accounting
In large businesses, many entities purchase a variety of products from many vendors. In these very complex systems, tracking activity that is typical versus fraudulent or wasteful is made difficult by the large number of transactions that are poorly documented, lack metadata, or have data entry errors. Traditional approaches apply rules to filter and flag transactions that resemble prior fraudulent or wasteful activity. These deterministic methods fail to establish a normal behavior and therefore cannot identify any other anomalies.
A more sophisticated approach is able to understand baseline behaviors of business units or vendors and detect deviations. This requires the generation of metadata to understand expected behavior, identify and correct data entry errors, and establish similarities and relationships among business units and vendors.
Join us to learn how we used Levenshtein distance, Benford's law, social networks, and behavioral profiling to detect fraudulent activity.
Hulya Farinas, Senior Principal Data Scientist -- Pivotal
Sarah Aerni, Senior Data Scientist -- Pivotal
Tim Matteson, Co-Founder -- Data Science Central