Tuesday, September 15, 2009

Data Analytics to Detect Fraud

Some say that detecting fraud is fraud is like finding needle in a haystack. Often times this is true, but more often you don’t even know what the needle looks like or what haystack to look in. To overcome these obstacles data mining techniques can be used. One of the ways that is very powerful is data visualization. Using this technique you can “see” the anomalies much easier than just staring a list of numbers.

One of the earliest, but still powerful analytic like this is called “Benford’s Law” – also called “Digital Analysis”. The basic premise of this law is that certain leading digits in any random set of data will appear in a specific non-uniform manner or in a certain frequency. Anything that is outside that frequency indicates a non-compliant anomaly. For example, if an employee has a limit of approval of $5000, you might see a spike in the first two digits of “48” or “49” that is beyond what Benford’s law says it should be.

There are some great tools out there that can let you apply this to your data – including MS Excel (click here).

No comments: