Georgia Tech ISYE 6402: Time Series Analysis

I took ISYE 6402 in Spring 2026 alongside Machine Learning. Before enrolling I had seen a lot of negative feedback about the course, but I also noticed the staff were genuinely making efforts to improve it. A notable example was allowing Python for assignments, something that had not always been the case when the course was R only. I found that the course did not deserve much of the criticism it had received and could teach you a lot of useful concepts in time series analysis. That said, getting an A did require real effort.

Topics Covered

The course covers a progression of classical time series methods:

• Regression analysis — foundational statistical modeling of time-dependent data
• Univariate ARMA and ARIMA modeling — core classical models for stationary and non-stationary series
• GARCH and ARCH modeling — volatility modeling for heteroskedastic data
• Vector Autoregressive (VAR) models — multivariate time series modeling
• Forecasting, model identification, and diagnostics — applied throughout every unit

The final portion of the semester introduced more modern machine learning approaches to time series, though these were covered at a higher level and did not appear on exams.

Prerequisites

Prior coding knowledge is not strictly necessary since the course provides a class repo with code examples to work from, but having some familiarity with Python will make your life easier. The analysis assignments primarily used statsmodels and pandas.

Assignments

The course has two bi-weekly assignments that alternate throughout the semester:

• Multiple choice quiz — open book, covering recent lecture material
• Analysis assignment — apply what you have learned to real data, completable in R or Python

Basic setup code is provided, but the bulk of the implementation is up to you. The course is not very rigorous from a coding standpoint. What your grade really depends on is the quality of your analysis. Each assignment requires thorough written interpretation of your results, and that is where you will spend most of your time. In total the course required around 10-15 hours per week.

Group Project

The group project was one of the more interesting parts of the course but also one of the more frustrating. Coordinating with a team adds a layer of complexity that has nothing to do with time series analysis. A lot of other groups saw members drop mid-semester, which made things harder for everyone involved. My group stayed organized throughout and we finished well.

Align with your team early. The coordination overhead is real and can catch you off guard if you wait too long to get organized.

Exams

The exams are challenging. There are both multiple choice and analysis sections, and I found the multiple choice to be noticeably harder than the analysis portions. Both sections are open note, so going in well-prepared with organized notes makes a real difference. With enough preparation you can definitely do well, but don’t treat open note as a reason not to study.

Peer Review

One aspect of the course I was not a fan of was the peer review requirement. For each analysis assignment you are required to review three of your classmates’ submissions. In theory this is a good idea since you can learn from seeing how others approached the same problem. In practice, a lot of reviewers would take off points without providing any meaningful feedback. After spending several hours on an assignment it is frustrating to receive a docked score with no explanation of what was wrong or how to improve.

Grading

Your grade is primarily determined by the assignments, exams, and the group project. The assignments and final project are very achievable if you give yourself enough time to finish them properly, and near-perfect scores are realistic. The exams are where grades tend to take a hit, so going into them well-studied matters more here than in most courses.

Overall

ISYE 6402 is a solid course that covers classical time series methods in depth. I chose it as a free elective because I was genuinely interested in the subject, and it is also part of the OMS Analytics program at Georgia Tech, so it pulls from a broader student population than most OMSCS courses. It did not deserve the reputation it had in previous semesters, and the improvements the staff have made are noticeable. If you have any interest in forecasting, statistical modeling, or working with sequential data it is well worth taking. It is also a reasonable course to pair with another class, as long as that other course is not too demanding.