vlogƵ

Students

Machine Learning, Student-Powered

By
Lance Pauker
Posted
May 8, 2024
Pace Federal Reserve Challenge team pictured with Jerome Powell

Machine Learning. ItvlogƵs a buzzword of late, and for good reason. Effectively employing this type of artificial intelligence can help reveal otherwise unseen patterns in fields as diverse as weather forecasting, healthcare operations, and nearly everything in between.

Image
headshot of Dyson student Liam Chentoufi

ItvlogƵs becoming a big deal in economics, and the research of Dyson College student Liam Chentoufi vlogƵ25 exemplifies how when it comes to this sort of interdisciplinary learning, Pace is on the cutting edge.

Liam initially intended to major in business; until by happenstance, he attended a data-centric event hosted by the economics department during common hour. He cites this event as a huge turning point in his undergraduate careervlogƵwhich ultimately inspired him to change gears, and dive headfirst into the world of economics.

vlogƵMeeting the department and all of the professors has been transformative,vlogƵ says Liam. vlogƵItvlogƵs those little chances that let us find our niche. I couldvlogƵve said to myself vlogƵIvlogƵll just get lunch and go next year,vlogƵ but next year it might not have been the same.vlogƵ

vlogƵItvlogƵs those little chances that let us find our niche. I couldvlogƵve said to myself vlogƵIvlogƵll just get lunch and go next year,vlogƵ but next year it might not have been the same.vlogƵ

As a sophomore, Liam took Economics 590: Data Analytics (R and Python), a graduate course taught by Professor Mary Kaltenberg, PhD, that merges economics principles with coding languages to make better use of data. As part of the class, Liam was tasked with a data analysis project.

vlogƵMy original question was trying to test if therevlogƵs any relationship between the locations where the Federal Reserve gives their speeches and some economic variable. Professor Kaltenberg mentioned at one point that this data is really rich for machine learning because itvlogƵs textual, but I wouldnvlogƵt be able to do it for my 590 class because it was too long a project for a single semester.vlogƵ

The seeds now planted for a more involved research study, Liam applied for and was accepted to the Office of the ProvostvlogƵs Undergraduate Student Faculty Research program. With the mentorship of Professor Kaltenberg, he began to investigate how machine learning can be applied to better understand and predict future decisions from the Federal Reserve.

vlogƵThe Fed, they conduct our nationvlogƵs monetary policy. When making their policy decisions they communicate with the public through these speeches that they publish every week or so. This project is trying to predict what theyvlogƵre going to do at their next meeting based on what theyvlogƵre saying in their speeches today,vlogƵ he notes.

Each Federal Reserve speech is publicly available, so LiamvlogƵs research involves scraping the text, and building a machine learning model to glean the attitude that is expressed in the text. HevlogƵs currently experimenting with two different machine learning architectures; one, a simpler model that requires less computing power but has 60% accuracy in classifying the text. The second, a more complex model that requires computing power beyond his personal laptop, is an adaptation of a model called Roberta developed and honed by Google and Meta.

vlogƵ[RobertavlogƵs] a larger model thatvlogƵs more complex, itvlogƵs harder for me to understand but it does yield more accurate results; it takes more time and computing power to train the model,vlogƵ explains Liam.

Professor Kaltenberg, who has been advising Liam as to the types of machine learning tools and resources he can utilize and experiment with for the project, believes this research is very representative of where the field is going; and is proud of the fact that Pace is among one of the first institutions to explore this unique research question.

vlogƵEmbedding machine learning within economics is increasingly important, and itvlogƵs one of the reasons why we created a new degree, computational economics,vlogƵ says Kaltenberg. vlogƵLiamvlogƵs project is exactly a reflection of that trajectory within economics, and across many different industries. This particular topic, trying to predict what the Fed will do next, can be extremely lucrative. We are not the first to consider it, but wevlogƵre among the first.vlogƵ

This particular topic, trying to predict what the Fed will do next, can be extremely lucrative. We are not the first to consider it, but wevlogƵre among the first.

The research dovetails very nicely with LiamvlogƵs future plansvlogƵthis summer, hevlogƵs landed an internship with the Federal Reserve Bank of Boston, where he will further learn about the machinations of monetary policy. After the summer, he plans on deciding whether to pursue a graduate degree in economics or enter the workforce upon graduation in spring 2025.

Through this research, his strong academic achievement, and his interests outside the classroomvlogƵhe is co-captain of PacevlogƵs nationally recognized Federal Reserve Challenge teamvlogƵLiam is setting himself up for future success no matter what route he takes. But perhaps most impressive, is the attitude he has developed in part through this research; one that has helped him develop into a more resilient individual, more confidently able to tackle whatever may be thrown is way.

vlogƵYouvlogƵre learning and testing things on the fly, itvlogƵs taught me a lot about myself,vlogƵ says Liam. vlogƵWith coding, you can receive so many errors, which is head-bangingly frustrating. But when it works, it is euphoric.vlogƵ

vlogƵThatvlogƵs the biggest thing,vlogƵ he adds. vlogƵLearning how to push through hard things.vlogƵ

More from Pace

Students

PacevlogƵs new Computational Intelligence Lab is officially open, serving as a hub for those interested in improving their programming skills, learning more about pattern recognition and artificial intelligence, and finding a place for like-minded people to congregate and collaborate.

Pace Magazine

Meet Christelle Scharff, PhD, a computer science expert focusing on the limitations and biases of AI systems. She and her team are tackling the intersection of AI and African fashion to explore the impact of diverse datasets.

Students

From immersive research to on-the-ground experiences with coopers in Spain, the PaceDocs team is once again crafting a captivating narrative. With dedicated translation efforts, meticulous scriptwriting, and seamless editing, the team navigates challenges to produce a stellar final product which will premiere on May 2.