AI-Assisted Financial Analysis
Spring 2024
Instructor
Kerry Back
(kerryback@gmail.com)
J. Howard Creekmore Professor of Finance and Professor of Economics
Meeting Schedule
Room 318 McNair Hall
TTh 12:30 – 2:00
3/12/2024 – 4/18/2024
Course Description
Each of us will, in the future if not already, use large language models (LLMs) as assistants. The models and interfaces will change over time, but it is worthwhile to learn what is available now and adapt as things evolve. Undoubtedly, everyone has already used ChatGPT and other LLMs for some purposes. What we will explore in this course is using an LLM with python to perform data analysis, specifically financial data analysis. We will use Julius.AI, which combines access to several LLMs (including ChatGPT4) with a python interpreter (which I’ll call LLM + Python).
At the end of the course, I anticipate that everyone may still prefer to use Excel for most things. However, there are some things for which an LLM + Python is better suited, and there are other things for which an LLM + Python may be faster. In any case, it is worthwhile to learn what can be done with this new tool, so we will be better prepared for the future. It is quite possible that everyone will end up using Python in Excel and Microsoft Copilot, rather than Julius.AI, but Python in Excel is not generally available yet, and, in any case, Microsoft Copilot + Python in Excel will also be an LLM + Python (+ spreadsheet).
The format of the course is that we will successively consider different types of analyses. We will perform each analysis live in class using Julius.AI. Then, your assignments will be to have similar conversations with Julius.
We will spend roughly a third of the course on topics that were covered in Applied Finance. This course will differ from Applied Finance in that we will ask an LLM + Python to do the work for us, instead of using Excel and @Risk. In the other 2/3 of the course, we will cover new topics, including machine learning and forecasting.
Topics
- Simulation
- Data handling
- Cost of capital and performance evaluation
- Visualization
- Portfolio optimization
- Autocorrelation and autoregression
- Machine learning
- Overfitting, shrinkage, and linear models
- Random forests and gradient boosting (application = valuing houses)
- Classification (application = predicting loan default)
- Neural networks
Grading
Grades will be based on weekly individual homework assignments (60%) and a final exam during exam week (40%). The final exam will be a timed at-home exam distributed through Canvas.
Julius
Julius.AI provides a 50% academic discount. Sign up for a free account, then send an email using your Rice email account to team@julius.ai and ask for the academic discount. They will respond with a promo code to use. The Basic account allows 250 messages per month and will probably be ok. If the message limit becomes binding, you can always switch at that time to the Essential account, which allows unlimited messages. Everything is run in the cloud from a web browser, so there is no software to download. In particular, it is not necessary to have python installed on your computer.
Honor Code
The Rice University honor code applies to all work in this course. Each student must do his or her own assignments, but it is allowed and in fact encouraged for students to seek advice from each other.
Disability Accommodations
Any student with a documented disability requiring accommodations in this course is encouraged to contact me outside of class. All discussions will remain confidential. Any adjustments or accommodations regarding assignments or the final exam must be made in advance. Students with disabilities should also contact Disability Support Services in the Allen Center.