« Back to blog index

Fighting Fraud with Big Data Analytics on Hadoop

Fraud is especially rampant this time of year and it grows every year. Retailers expect to make up to 40% of their revenue for the year during the holiday season. Yet last year e-tailers lost $3.5B to online fraud. And retailers are not alone. Recent studies have found merchants paying $200B to $250B in fraud losses annually. Banks and financial organizations are losing $12B to $15B annually. Exacerbating the issue is the high data volumes—over 20B for credit card transactions are made annually.

And the face of fraud is changing.  Instead of stealing a credit card and using it to buy big ticket items like big screen TVs, credit card thieves have become more sophisticated. For example, they can now make numerous, small transactions that are seemingly benign. But if Joe is making 100 $5 coffee transactions at multiple locations at the same time, something is wrong. By analyzing large volumes of complex data—including point of sale, geolocation, authorization, and transaction data– with Datameer on Hadoop, companies were able to identify fraud patterns in historical data.

The reference architecture required to support fraud detection in this new world needs to support business-user focused big data analytics applications on top of a Hadoop based architecture.  See here for a demonstration on analytics used to detect fraud including:

  • identifying outlying spend and affected vendors
  • data mining and machine learning on transaction data
  • predictive modeling (e.g. back-propagation)

Karen Hsu

Karen Hsu is Senior Director of Product Marketing in Datameer, Inc.

Comments are closed.