Follow Eccentex

Subscribe to the Eccentex Newsletter

Your email:

    

Dynamic Case Management Digest

Current Articles | RSS Feed RSS Feed

Investigative Data Sources: Getting Bigger and Badder

  
  
  

Guest Post by Dr. Manuel Aparicio of Saffron Technology

big dataBig Data challenges are effecting every industry, but what does it mean for fraud detection and investigation? First, Big Data itself is irrelevant; the real challenges are in Big Data Analytics, in how to exploit the greater volume, velocity, variety, and volatility of data - for greater value. Two of these Vs of Big Data are particularly relevant to the growing requirements of fraud detection and investigative processes.

Variety. Data is getting bigger in part because of the growing variety of sources. To address Anti-Money Laundering for example, a bank will have any number of different transactional systems for different transaction types along with a number of know-your-customer databases. It might also purchase public figure and associates, aliases and watch list databases, not to mention all sorts of other external data sources that add more knowledge to detection and investigation. These sources all have different schemas and may be structured, unstructured, or both. New "schema-free" approaches from the NOSQL ("Not Only SQL") movement are required to address the number and variety of such data sources. In particular, if investigators are always trying to "connect the dots," then graph-oriented knowledge representation, which connect the dots between people, places, things, events, etc are required to help by doing the same.

Volatility. The dimension of volatility is even more interesting. In today's world, data is always changing. This means more than saying that new data is constantly arriving, which is defined as velocity. Volatility means that the world itself, it's underlying model, is always changing. In the context of fraud detection, the fraudsters are always adapting, always changing strategies and behaviors to defeat the detectors. They are getting "badder" in the sense that fixed detection models trigger on previously known "bad" patterns but start to break when the patterns change to something new or slightly different: when fraudsters get badder than models can detect. But why can't the models of bad behavior - the detectors, learn on the fly when a new pattern is discovered and adjudicated? This is a problem of "models," based on traditional rules and statistics, requiring data scientists to develop, maintain, and update them, which requires ongoing cost and time to keep up with the volatility. This is where a new breed of machine learning is now coming into play and dynamic case management applications come to the rescue. Similar in philosophy to "schema-free" graphs, "model free" statistics (sometimes called "nonfunctional" modeling) is emerging to address what is increasingly required of fraud detection systems: instant learning of new patterns, from investigation to adjudication back to detection.

The mind of the investigator works this way: always connecting the dots and always learning from experience. A new breed of technology thinks the same way. These new dynamic approaches are required as data gets bigger and badder.

b77fbdd9-3b7c-4a7a-9c1c-0999a1187435

 

manuel aparicio

Dr. Manuel Aparicio is a Chief Executive Officer and Co-Founder of Saffron Technology where he sets the company’s corporate vision and fundamental direction. With more than 25 years experience in the industrial development and commercialization of software solutions supporting real intelligence for intelligent agents, Aparicio has long been an evangelist for broad adoption of associative-memory technology, or what Saffron calls Experience Management solutions. Aparicio holds several patents for memory-based technology, and has published numerous papers and journal articles, such as “Concepts of Personalization” in The Practical Handbook of Internet Computing, and “Learning by Collaborative and Individual-based Recommendation Agents” in the Journal of Consumer Psychology. Aparicio formed Saffron in 1999, and currently serves as a technology advisor to Sociocast, LLC, a Web 3.0 social influence network company.

Rapidly build and deploy Dynamic Case Management Applications in the Cloud with AppBase, Eccentex's Platform-as-a-Service. Visit www.eccentex.com to learn how to increase business agility, improve productivity, optimize case outcomes, and reduce risk with a cloud platform.

Comments

Currently, there are no comments. Be the first to post one!
Post Comment
Name
 *
Email
 *
Website (optional)
Comment
 *

Allowed tags: <a> link, <b> bold, <i> italics