Introduction to Analytics Modeling

Dreadful, actually pathetic that a top 5 school allows this course. The lectures too sparsely cover too many models, the homework are LLM coding exercises with no feedback beyond thumbs-up emojis or trivial formatting criticisms, and the exams are largely indecipherable.

If you want a basic introduction to ML, statquest is an order of magnitude better than 6501 and 100% free.

I am astounded this class is rated so high and chalk it up to people with CS undergrads or people with low standards. This course is so bad but rated so high I withdrew from the entire program. Most of the other required courses are in the low 2s. Shocking, because I can't even imagine how bad they must be after taking this one.

TL; DR for ISYE 6501: Interesting lecture material + homework (15%) and project (8%) with work potentially sabotaged by peer grading. Horrible exams (75%). Some of the exam questions were so oddly worded that if English was not your native language you were at a distinct disadvantage. Enrollment: 1300+ students! Class subject matter+ HW assignments + project: 5/5 Peer grading and exams: 1/5 Overall course rating: 3/5. Exams and peer grading are red flags.

Grades: HW + Project 100, Exams (average) 87, Final grade A (90%)

Pros (Top 3):

1. Excellent choice of topics and covered in just enough detail by Dr. Sokol in well-organized lectures.
2. Interesting homework assignments, designed with the lectures in mind. HW consists of 80% of your time in this class and where you will put work in. This course needs to use the auto-grader for the coding.
3. The project assignment allows a deep dive into a particular problem and requires you use at least three of the methods acquired in the first 10 weeks. As with the homework you get out of the project what you put in.

Cons (Bottom 3):

1. Exams: 75% of Grade. It’s a shame that the often ambiguous and poorly worded exams account for 75% of the grade in this course. Many exam questions introduce an unnecessary level of ambiguity that is unsettling.
2. HW: Homework and Project (23% of Grade) are all peer graded, and account for almost a quarter of the grade. Some of the peer reviews I received were performed with only a modicum of effort if any. Given the quality of gatech.edu as an engineering school, the lack of an auto-grader for coding exercises is unforgivable.
3. Office Hours (OH): Due to work pressures I had to finish the HW well before OH on Monday. So, OH were not very useful. As reported in OMS reviews, some students waited for the office hours the Monday before the HW was due and then copied the work done there. Some of the HW rubrics were weak and nearly incomplete.

Six course design fixes that would allow ISYE 6501 to achieve a 5/5 rating:

1. Exams: Make exams better reflect the material not the ability to dissect triple negatives and split infinitives.
2. Grading: Please reward work on homework writeups. At minimum a 50%-50% split between HW/Project and Exams.
3. Coding: Include a 5-10-minute TA-led segment demonstrating R or python code relevant to HW as part of the lecture.
4. HW: Implementation should require the student to deep dive the details. Provide more test problems.
5. Require 2+ of the HWs (chosen randomly) to have one TA grade to ensure quality grading & identify poor peer graders.
6. Get rid of Piazza, use Ed.

I wanted to share some honest feedback after taking two midterms of ISYE 6501. I’ve really enjoyed the lectures and learned a lot from the assignments — even though they’re quite time-consuming, they’ve been deeply educational and practical. But I feel that the exam setup and grading structure don’t really align with the spirit or learning goals of the course.

To start with, the exam questions feel misleading and unevenly weighted. Some questions are worth 5–6 points while others are only 1 point, and it doesn’t seem that this distribution was carefully reviewed. In many cases, it felt like the challenge was not about applying what we learned, but about trying to interpret what the question writer was actually asking. That creates unnecessary stress and confusion, unrelated to the actual course material.

I can’t help but wonder — are the exams thoroughly reviewed by Prof before release? Because the quality of the lectures is excellent, but the exam feels disconnected from that level of care.

Also, I think the grading structure overall needs reconsideration. Homework only counts for 15% of the grade, yet it easily represents 60–70% of the actual learning in this course. I understand it might be weighted lower because it’s peer-graded and perhaps not taken as seriously in scoring, but that’s exactly the problem — it’s where most of the effort and meaningful understanding happen. If that’s the case, then the peer-grading system itself should be reevaluated, not just for the sake of TA workload, but to ensure that students are fairly rewarded for the depth of learning these assignments provide.

Meanwhile, exams make up 75% of the grade, even though they often don’t reflect how well someone understands the material — just how well they can navigate ambiguous or oddly worded questions.

I truly think ISYE 6501 is a great course with strong fundamentals, but it could be so much better if the exams and grade weighting were revisited to better reflect the effort and real learning happening throughout the semester.

I had mixed feelings about this course.

On the positive side, I found the course content and rigour to be a solid way to kickstart my OMSA journey. The professor is super engaging and enthusiastic, and his lectures explain both the theory and examples very well. That said, the homework is largely self-directed - you’re expected to code independently (+ learn to code for those with no R background). The first homework was by far the most challenging for me (and possibly for other students too), which caught me off guard. After that, the remaining assignments were more manageable. Just note that there are one or two of the later weeks that need Python, which can be tricky if you don’t have prior experience.

Peer grading was frustrating at times since it’s inherently subjective - often the difference between a 90% and 100% depends on the grader, and many seemed to approach it as a chore rather than a peer learning opportunity. Personally, I would have preferred more (or any) TA grading, though given that homework makes up only a small portion of the final grade, it’s hard to complain much. There’s also a mini project, but it’s fairly light - no coding required, just a short report of your problem/use cases and proposed analytics.

The quizzes are what really determine your final grade. As long as you keep up with the lectures, you should be fine, though some of the questions are worded in a tricky way. Unfortunately, the quizzes don’t tie in with the homework at all, which further adds to my impression that the course workload feels disjointed.

TLDR: Great professor and lectures, but homework doesn't really align with the effort required (so don’t stress over it too much).

As an OMSCS student this was a great course to take and is basically a statistical learning / ML 101 course. It doesn't dive too deep into the math behind any models but will build a good foundation of how to approach modeling and what models are appropriate for what scenarios.

The lectures were a 10/10 and extremely informative, the homework could take a while but were really good for solidifying information. Although homework was only 15% of the grade and is peer reviewed (a 90 is for a complete homework, 100 is only if you go above and beyond). I found that even if I went above and beyond sometimes I'd get graded a 90 and other times a 100 so it's a toss up.

The course is exam heavy (75% of your grade across 3 exams) and 95% of the information is from the lectures. The median grades were low to mid 80s and I found t exams to be fairly challenging. Some of the wording for the questions threw me off but you get used to it after the first exam.

My advice would be to just do the homework to completion and put the rest of your effort into learning the material for the exams. There also seems to be a small unofficial curve (I got an A with an 89% actual grade). This course is an easy B if you do the work and have decent exam grades. To get an A you really need to crush the exams or get an 87+ and hope for a curve.

Overall it was a great course to take, just be prepared to get a B if you struggle on the exams.

I liked the course but it's not as easy as some people make it out to be. Professor Sokol is very knowledgeable and following the lecture content is rarely dry. The videos are well-made and the material is important for your analytics career as you cover important topics like regression, principal component analysis, and optimization to name a few. This is an introductory course so don't expect a full in-depth review. It's just enough to get your feet wet.

Homeworks (15%), 1 project (8%), and intro survey (2%) take 25% of grade while the remaining 3 exams take 75%. You will need to learn R for the homework's which can be time consuming in the beginning. 2 homework's are dropped so do well on the earlier assignments. They are graded by your peers so it's pretty difficult to get 100's unless you go above and beyond which for 15% doesn't really seem worth it. I ended up doing well (mostly 90s on the beginning homework's) and just didn't do the last homework's regarding optimization which was the last topic of the course. The project is just a super simple report on an analytics case study you choose (no coding required).

Now we get to the exams. I'm generally a decent test taker but these exams are worded in such an unusual way that it almost throws you off guard. Not even the content of the questions but the phrasing. Multiple answers may fit but the best answer is needed which can be confusing well. I think the mean for all the exams was high 70s, low 80's. The exams are my only gripe with the course, I wish they were a little more traditional. The homework's are not needed for the exams so in theory you could ignore every single homework and still pass if you do well on the exams (but don't do this lol).

Disclaimer: the class also uses Piazza (I think all the OMSA use it but I could be mistaken) and I rarely went on it.

I'm in the CS track and I'm glad i took this course. One of the better courses to take given the importance of its content and how it was presented just don't expect an easy A. An easy C, a B with decent effort, and an A if you do really well on the exams.

The Good: The material was interesting and the lectures were pretty good. With a background in analytics, the material was not extremely difficult but still provided some challenge at points. There are plenty of TAs, most of whom were extremely helpful. Most Piazza posts were answered promptly and well. The office hours were extremely helpful in completing the homework assignments.

The Bad: The structure of the course is a little weird at points, but my major gripe with this class was the exams. The homeworks do absolutely nothing to prepare you for the exams, so don't bother with reviewing those to study. They are also worth so little of your grade compared to the amount of time you spend on them that it almost feels like a waste of time. The exams make up 75% of your grade, which wouldn't be as big of a deal if the exams were written better. Some of the question wording in the exams is intentionally confusing, trying to force you to make a mistake. I never understood why exams would contain "Gotcha" questions that don't actually test your understanding of the material. The second exam was the most clearly written, but midterm 1 and the final were more confusing. The first two exams contain essentially no questions regarding code, but the final randomly has questions about R functions. They also like to have true/false questions that aren't formatted as true/false (you are given a statement with a missing word and have to choose from 2 options like COULD/COULD NOT). Additionally, the question count on the exam is a little misleading as they love to have several questions with 6+ parts to them, especially in the final. I also felt that Dr. Sokol was relatively absent from the courses as it's clear the lectures were recorded years ago and he does not attend most if any of the office hours. Another major gripe is having all assignments and the project be peer graded. This leads to a lot of inconsistency in the grades for these. The grading scale is a bit annoying (getting everything correct only gets you a 90%, to get a 100% you are supposed to provide a "Deeper solution than expected". Some weeks I put in a ton of effort and got a 90%, other weeks I put in very little effort and still got a 90% or one time an 100%. Some students oddly get a bit of a power trip on the peer grading and nitpick the tiniest things to take points away (which is honestly a little sad). One student gave me a 75% on an assignment saying I got everything right but they didn't like the way I formatted my assignment.

Overall, it's a fine class, but I wish it was better especially given it is a core class. I felt like most of the reviews I read pretty well summed up the course. I think it could benefit from having a bit more project based work, with something similar to CSE 6040 having autograded coding assignments and exams.

I came into this course with very little R experience or knowledge, so the ramp in the beginning was pretty steep. While the initial shock softens after 2-3 weeks, the class is by no means easy, but the effort and time are worth the volume of useful information you learn. If you want to make the most of the course, use the ISLR as a supplement and make the most of the homework assignments - experiment & go beyond the assignment. It's worth it.

I've read some reviews complaining about the wording on test questions. I think they're mostly correct: some of the questions do seem to be worded/structured to create confusion. Still - it's Georgia Tech. Every class you take is going to try to make sure you're thinking, learning and applying the info, not just regurgitating facts. They don't just give away Master's degrees.