Recently my office announced that we’d used advanced machine learning to find someone who’d illegally obtained unemployment compensation. We’re calling our use of data analytics to find unemployment fraud “Operation Payback.”

Use of the most advanced technology is going to be critical to uncovering more fraud, waste, and abuse in the future. My office has already used this new big data technology in multiple ways to help the taxpayers.

For example, data analysts in my office looked at millions of data points from Medicaid a couple of years ago. We not only found that some on Medicaid looked to be ineligible for the program, but also that some on the rolls were dead. When we turned that information over to the state Medicaid program, they were able to save taxpayers around $600,000 for the year.

Another advanced data analytics project we’ve done involves one-time vendors for state government. In Mississippi, if a vendor performs only one service for state government—let’s say a plumber comes by once and fixes a pipe at a state agency—then that vendor is subject to far less red tape than a vendor who performs hundreds of services for the state. However, there’s a problem if a vendor performs many services but pretends they’ve only performed that service once so they can avoid paperwork. They might be avoiding the paperwork so they can conceal fraud, like a kickback to a government employee.

There are many techniques these vendors can use to conceal performing multiple services. A fraudulent vendor could spell their name slightly differently (Shad White vs. Shad T. White vs. Shadrack White) each time they perform work. Given the tricky ways that some vendors try to keep this work concealed along with the massive number of transactions in state government, catching folks who want to take advantage of the one-time vendor rules can be difficult.

So my office used artificial intelligence (advanced software) to develop a method to identify which of the one-time vendors were not actually one-time vendors.

Finally, my office partnered with an advanced data analytics firm for a performance audit of six school districts. This project wasn’t looking for fraud, necessarily, but was simply making sure that more of our money makes it into the classroom. Data analysts were able to look at every expenditure made in a school district and analyze them for inefficiency.

Using data provided by the school district, we found numerous ways each school district could save money. These recommendations included eliminating underutilized software programs, employing a district-wide purchasing officer, bringing transportation costs in line with regional averages, cutting administrative salaries, and eliminating unused buildings.

And we’re not stopping. We’ve got plans to continue to use advanced data analysis to find fraud in healthcare and other places where government spends a lot of taxpayer money. If you’re thinking of defrauding Mississippi taxpayers you should understand that we’re not only watching, but we’re using machine learning to find needles in haystacks. You’re on notice.

 

Shad White is the 42^nd State Auditor of Mississippi