Students Apply Machine Learning Skills to Summer Finance Course
By Aubrey Buberniak
The summer of 2020 may be coming to an end, but several Carnegie Mellon students will be ahead of the game when they return to classes in the fall. Though many internships and co-ops were canceled or delayed due to the coronavirus pandemic, students found creative ways to continue enriching their education by participating in the university’s Experiential Learning Project courses — a summer school of experiential learning and remote research opportunities, all at no additional cost to students.
Bryan Routledge, Associate Professor of Finance, and Burton Hollifield, Professor of Financial Economics, PNC Professor of Finance, and head of the Undergraduate Business Administration program, designed an experiential learning project course, titled “Machine Learning and Finance.” Roughly thirty students from throughout Carnegie Mellon participated in the course, which ran from June 1 to July 15.
Data-Based Decisions
Students were strategically organized into teams of five, all with varying levels of experience and education. Teams were then tasked with identifying a project focus based on analyzing scalable sets of data variables and choosing methods to then create financial predictions.
The topics ranged from financial portfolio management during a pandemic to the impact of news and social platform trends on market values to the liquidity in the municipal bond market.
Many of the students who participated in the course couldn't pass up the opportunity to gain exposure through experiential learning.
“I have a strong interest in applying data and technology to finance, and although I have some academic knowledge in these areas, I was eager to use my education to solve real-life problems,” said Tracy Wang, a senior at the Tepper School.
“My main goal was to learn as much as I can while working with different types of data, algorithms, and models, and a hands-on approach to learning was an effective way to continue to build this knowledge.”
The course was designed to be flexible. Students pitched their project focus, then met with advisors and professors several times throughout the class for guidance, but ultimately worked independently while analyzing their findings and creating portfolio and investment strategies using machine learning approaches.
“I was particularly struck with how nice of a job the students did at taking ownership of their projects, the techniques that were used, and ultimately, in the variety of projects that were presented,” said Routledge.
“We gave them broad parameters on data, narrowed the focus to have a finance implication, and the students took these datasets and their interests, and created both portfolio-related tasks and different machine learning techniques to help drive through a business or economic decision.”
Applying Hard Skills Learned in the Classroom
For many of the students, the machine learning course was an opportunity to take the variety of skills that they learned in the classroom and apply them in practicum.
“I was able to apply skills from many different courses and lab work while taking this class,” said Kyle Bannerman, a College of Engineering senior.
“Whether it was setting up a systematic data pipeline, cleaning the data, or actually making predictions on the data, there wasn't a single course that prepared me for all of this but were skills I had developed over my time at CMU. Some of us in the group had never worked on a finance problem before, yet we were able to use our machine learning and data science backgrounds to create a profitable trading strategy under a relatively short time constraint.”
“Generally, CMU’s focus on analytics-based teaching and interdisciplinary education played a key role in the success of this project,” said Raahil Reddy, a senior at the Tepper School. “It was only because of the interdisciplinary education that we were able to combine technology, statistics, and finance in our project.”
Bryan Nowlan, a junior at the Tepper School, agreed. His previous classes on investment analysis and finance helped tremendously in the course.
“My coursework was paramount in understanding the model outputs and hypothesizing why certain variables were influential and others weren't,” he said. “Without my time at CMU, we would have been at a loss on how to even approach the project.”
Developing Team Strategies for Future Work
Putting students from varied backgrounds together on an integrated team led to more than just a diverse collection of educational skills and experience levels; students had the opportunity to work on their communication and leadership skills as well, especially considering that the course was completed remotely.
“Our project took a considerable amount of communication and teamwork, which were skills that had been stressed in the business school for my entire career,” said Jonty Nobbs, a junior at the Tepper School.
“Sure, I learned a whole load on how to code in python using pandas and sci-kit-learn, but realistically, the skill that will stick is communication and determination.”
As business needs change and technology continues to be developed, courses such as this become a foundation for a path forward.
“My final takeaway from the project is that data-driven methods are going to be the future of finance,” said Reddy.
The students’ experiences will prepare them well for their future careers.
“To me, the course brought together many of the elements I like about working with students at Tepper and CMU,” said Hollifield.
"I am impressed in the students’ ability to work in multi-disciplinary teams, as well as their intellectual curiosity, creativity, and energy, and their ability to bring business problems, analytical approaches, and technical skills together in solving real-world problems. Their projects are at the frontier in combining problems in finance and analytics."