Subscribe to DSC Newsletter

Predictive Analytics for Financial Services

The Federal Trade Commission’s Consumer Sentinel Network monitors the history of fraud and identity theft complaints. Their report on calendar year 2012 highlights the 2,061,495 cases of fraud, identity theft, and other infractions that were reported to the organization from every state, with Florida, Georgia and California leading with the most reports. That’s over 2 million cases of fraud and theft that were reported, but what of the ones that went undetected? While some individuals, business, and organizations were vigilant enough to notice fraud as it happened, many others either don’t know where to begin looking or are being hustled by lambs in sheep’s clothing.

Predictive analytics could be helpful for understanding the patterns of people who commit this type of crime. With the number of cases of fraud and theft occurring each year rising significantly, though, predictive analytics needs a helping hand. Fraud and identity theft are occurring in increasingly creative, sneaky ways, many of which involve hiding actions behind different names or other identity tags. Identity fraud not only affects the customer, it also affects that person’s bank and other vendors when identity theft strikes. The imperative to stop identity theft and fraud is significant on both individual and systemic levels.

Download the Big Data for Banks ebook for free here.

How does a bank or credit-scoring institution find these outliers when they deal with an incredible amount of information to begin with? This mass of data requires automatic pattern detection. When one business or individual begins to refer to themselves under multiple names or identifiers, the trail of their actions can easily be lost on human analysts. Factor in the compounding ways and speed through which identity theft can be committed, and the potential for human analysis as a means of detecting patterns seems foolhardy. Pattern discovery is essential when looking for the intricacies of fraudulent behavior.

A human investigator might have a brilliant mind, but the types of patterns that add up to uncovering identity theft today are simply too complex in fact and size for a bias-prone human to uncover—let alone the fact that some identity theft occurs across wildly disparate parameters that humans might never connect simply because at first glance the categories do not overlap. At the end of the day, automatic pattern detection might uncover a fraudulent actor, or it might uncover the pattern of a person who looks fraudulent, but is actually a poor typist or moves around a lot. Either way, the patterns of behavior and commonality across entries are objectively and automatically located regardless of names listed, leading human analysts towards the insight needed to detect, stop, and prosecute those who are stealing identities.

Download the Big Data for Banks ebook for free here.

Almost every player in every industry now has to accept credit cards due to user demand. As the transfer of money becomes increasingly more digitized, the chance to compromise points of transaction increase with every new point of transaction. With the advent of technologies like Square and other services that allow users to accept payments anywhere, the number of credit transaction sites has exploded. The more transaction points, the greater the chance to fraudulently use credit cards and go mostly unnoticed. This is why banks and credit processors need a tool to very quickly and accurately target patterns in use that indicate potential fraud. The more fraud happens across multiple transaction points, the larger field of potential harm to businesses and other groups who conduct transactions via credit cards. A wider field means an infinitely larger amount of incoming data.

When data bubbles to the size and speed of credit card transactions, human analysis can be completely forgotten about as a means for discovery of patterns. There are too many dimensions for teams of even tens of people to consider discovering the patterns that reveal the credit card fraud across transactions. Factors that indicate potential credit card fraud expand well beyond the transaction information itself—they include the time a transaction occurred, the location of the suspect transaction or transactions, the day of the week they occurred on, and so on. Consider big-box retailers like Walmart that see millions of people across thousands of stores a day. The number of transactions across that one set of retail storefronts is too much to handle alone.

Download the Big Data for Banks ebook for free here.

Just as with other general types of fraud like identity theft, fraudulent credit users attempt to cover their tracks, or seek out friends and acquaintances at specific retailers who are willing to collude with their theft. By automatically detecting patterns in transaction data that indicate something fishy, analysts can free their time away from virtually impossible manual pattern detection and focus on a much smaller set of evidence. Less time to answers for analysts means minimized losses for businesses that are affected by credit card fraud. Fewer losses leads to a more robust economy at large, including lower prices for offered items.

By automatically eliminating the noise that surrounds the actionable insight in credit card transaction data, pattern detection enables analysts to ignore certain activity and focus on the truly novel, truly suspicious data patterns.

Views: 2337

Comment

You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service