Predictive analytics is used to spot missing and suspicious payments for agencies
Agencies that are just getting started with artificial intelligence and predictive analytics may face internal inertia. In this case, starting small can help.
That’s according to Michael Collins, a statistician at the Federal Student Aid Bureau of the Department of Education. Speaking at the agency’s predictive analytics readiness panel on Tuesday, presented by the Advanced Technology Academic Research Center, Collins described how his office, which administers FAFSA form, uses predictive analytics to predict default.
Now that the federal student loan remittance of payment, arrests of collection and waived interest – due to COVID-19 – are should end on December 31, FSA has started targeting its communications to borrowers who they believe are most vulnerable to defaulting on their loans. But the agency also performs this predictive modeling at several stages of the loan life cycle.
“We will predict this when a candidate first files their FAFSA form and asks for help, and we will do that again at a time when they are actually in college and their aid is being paid. And then we’ll model it again, when they’re out of school, graduated, or dropped out, and model a probability of default at that point, ”Collins said. “One of the most powerful predictors of whether a borrower will default on their loan is whether they have completed the program and graduated. And we won’t know until they finish or not, until they refund.
FSA can use this probability of default to forecast future cash flows, plan outreach, and test messages. Borrowers may not be aware of resources and repayment options, he added.
“We’ll have borrowers who just forgot, as they maybe have a six-month grace period after they finish school to start payment. And they just didn’t. And it will clearly be a very different situation than someone who has been making payments for five years and then suddenly stops making payments, ”Collins said. “So someone starts to be late with their payments and they haven’t defaulted yet – we want to help them avoid any default. “
Predictive analytics can also help detect internal failures.
Chakib Chraibi, Chief Data Scientist at The Commerce Department’s national technical information service said its agency uses predictive analytics in conjunction with organizations to optimize resources. This helps mitigate risk by proactively shifting resources to areas where something could likely go wrong, he said.
Most often, predictive analytics is used in forecasting, whether with regression modeling, classification modeling, or other types.
In collaboration with the Office of the Inspector General of the Ministry of Health and Social Services, NTIS uses artificial intelligence and advanced analytics to spot suspicious transactions and put an end to inappropriate payments or fraudulent schemes. While detecting anomalies and overpayments, Chraibi recalled an anecdote of a nail salon that billed Medicare like surgery.
“Look at the models – and this is a difficult area in predictive analytics. We must therefore proceed step by step, ”he said. “First you need to collect the data you need. And then you have to develop a system on how to spot anomalies and that includes subject matter experts who are there. “
Data visualizations are useful for spotting patterns that are “out of whack”. Data collection may also include public data from social media, news posts, and other sources.
And sometimes the NTIS has to work with agencies like the Ministry of Labor, where a lot of data resides in siled applications, making it difficult to perform effective predictive analytics, Chraibi said.
“We’re working on building a data infrastructure that will help them integrate all of their data and help each unit better understand what the data is telling them – not just within their unit, but overall. of the agency, ”he said. “We have therefore also made progress in handling errors, handling complaints, authenticating and verifying eligibility of applicants, and so on.