Mathematical models put food in mouths: Math student’s research aids local food banks

Food pantries have an inventory problem.

Each month, a food bank receives any number of visitors who are experiencing food insecurity. Each month, a food bank also receives donations that go to stock the shelves. Will the pantry receive enough donations to meet the demand of its clients? Will there be surplus, shelf-stable donations they can carry over into next month? Will there be an unusually high number of clients relying on the food pantry this month? How do you answer those questions?

“The best way might be to use a Markov Chain,” Gavin Sayrs mused.

Sayrs is an undergraduate research student at UWM. He is double-majoring in mathematical sciences and German, and he’s also the treasurer of the Living Hope Club, a campus organization that focuses on helping Milwaukee residents living below the poverty line.

“I thought, I really enjoy charity work, and I have a club that does charity work. Why don’t I take my interest in math and combine it and see what happens?” Sayrs said. “There’s food going in and out of pantries constantly. How can we best help them know, in the future months, how much food they’ll be receiving and how much food they might be losing?”

That was the inspiration for his undergraduate research project titled, “Local Food Pantry Hidden Markov Model Analysis.” He presented his work at the UWM Undergraduate Research Symposium in April. He was mentored by UWM mathematics professor Chao Zhu. Sayrs relied on data supplied by the Kinship Food Pantry in Milwaukee.

If you’ve never heard of a Hidden Markov Chain, you’re not alone, but it’s a tool that’s being increasingly used in supply-chain and logistics fields. Think of the Markov chain like a matrix with many probabilities within it. Each matrix entry shows where the probability will go over time. You can use a Markov Chain to predict a sequence of unknown (hidden) variables based on a set of observed variables.

“To summarize: Every single time a transition is made from an unobserved state to another unobserved state (it could possibly be the same state it’s transitioning to), there is a signal emitted that can be seen that is the observed state,” Sayrs explained.

“In our situation, the observed state is how much inflow or outflow of food we’re getting. That can either be a positive number where we’re receiving more food than giving out, or a negative number, where we’re giving out more food than we’re receiving.

“From there, we can work backwards. We’re only able to see data about how much food we’re receiving,” Sayrs added. “We can say what state we are in currently, and we can combine this data and say what (donation levels) will look like in the next year.”

In terms of inventory in a food pantry, Sayrs had plenty of observed variables, thanks to the meticulous records that Kinship Food Pantry keeps. Working with Krista Fuller, an intern at Kinship working as the Food Center Operations Coordinator, Sayrs gathered five years’ worth of data showing the food bank’s monthly donations and how much they served up to community members in need.

“We see that in certain months, we don’t receive as much food as in other months. There’s a certain inflow/outflow difference,” he said. “We can record this difference by taking all of our monthly data and looking at it through a standard deviation. Is our monthly inflow/outflow within one standard deviation, meaning everything’s normal? Or is it greater than one standard deviation? …From there, we can see how much food they’re getting in relation to what they normally get.”

With this method, Sayrs said, the food pantries can predict which months might be leaner than others. For example, September, October, and November are all months where the food pantry sees an increased number of visitors. The food pantry may want to hold back surplus food in July and August to make sure there is enough food on the shelves in the coming months, or workers might try to reserve money in their budget to address the shortfall.

There is still more research to do. While Sayrs is grateful that Kinship Food Pantry was generous in working with him (“They were so gracious,” he said. “They helped out all that they could and it was a great experience.”), he wants more than just the 5 years of data they were able to provide. To really build a good model, he said, Sayrs wants to look at 10 years’ worth of pantry inflow/outflow numbers.

Even so, Sayrs said he’s shared his finding with Kinship Food Pantry, and hopes that other food pantries in the city can also use this method to predict inventory surpluses and shortfalls. Hidden Markov Chains are already used in weather forecasting, estimating stock market prices, and even in projecting the spread of diseases.

“I know people who get involved in working with food pantries might not have math backgrounds. They have some social science background or something like that. That’s good. We need people like that,” Sayrs said. “But we also need to have solid data, because that helps them do what they’re best at doing.”

By Sarah Vickery, College of Letters & Science


Share:

UWM Land Acknowledgement: We acknowledge in Milwaukee that we are on traditional Potawatomi, Ho-Chunk and Menominee homeland along the southwest shores of Michigami, North America’s largest system of freshwater lakes, where the Milwaukee, Menominee and Kinnickinnic rivers meet and the people of Wisconsin’s sovereign Anishinaabe, Ho-Chunk, Menominee, Oneida and Mohican nations remain present.   |   To learn more, visit the Electa Quinney Institute website.