SDSM&T students design software to predict the cattle market

RAPID CITY, S.D. — The South Dakota School of Mines and Technology (SDSM&T) Computer Science and Engineering Department is working on a project to design software that will predict the cattle market.

SDSM&T computer science majors, Jordan Baumeister, Dustin Reff, and Trevor Borman. Photo courtesy of SDSM&T.

The project uses artificial intelligence and historical data in the cattle and corn markets to create mathematical models that predict future market trends. The models take into account 187 different variables such as drought, disease, rainfall, price of hay and fuel prices.

The idea to create the model came from South Dakota rancher and mathematician Ron Ragsdale who worked with SDSM&T student Todd Gange in 1993.

“The rancher who had originally started liked this map in that model. He used it very wisely. He could basically tell whether or not it was worth him to lease his land or to actually buy cows. And I think that’s something that every farmer could use,” said Jordan Baumeister, a SDSM&T graduate who worked on the project last year as her senior project and is now a software developer for General Motors.

In 2021, Gagne shared the software that he developed with the SDSM&T sponsoring it as a Computer Science and Engineering Department senior project.  He challenged the students working on the project to enhance the program to better predict commodity prices when outside factors drive the market off its normal course.

Three students — Baumeister, Treavor Borman, and Dustin Reff — were the first team to work on the project last school year, where they weeded through nearly 50 years of historical data and developed two different computer models. The first uses the historical data to determine risk versus reward analysis and the second is a predictor model showing the best times to buy and sell.

“So what we have now doesn’t necessarily predict what the market will do, but we have a really good basis for taking that historical analysis and being able to use that just because the market trends seems to stay similar throughout each year. So you can kind of look at another year with this high inflation rate and look at how those contracts played out and they’ll probably collapse pretty similar to this year since it’s similar conditions,” said Baumeister.

Baumeister said that the first computer model using the historical data tested well.

“It only failed twice and that was during 9/11 and the Lehman Brothers collapse,” she said.

The second model that helps predict when to buy and sell cattle didn’t fare as well.

“Another teammate who is on the team, he did a lot of predictor analysis type stuff and he was just having a hard time trying to get something that could accurately predict that market,” said Baumeister. “I think the next phase is getting a predictive model that’s definitely more accurate because our predictions are terrible.” 

The project is ongoing and a new team will be working on the project this fall. The next team will be challenged to rebuild the model adding more historical data.

Gagne hopes the software will eventually be marketed, said Baumeister. She believes that is can be useful not only to commodity trader, but also to livestock producers, feed yards, and for meat processing plant procurement.