Reflections on Georgia Tech OMSA Program
Introduction
The purpose of this document is to provide an overview of my experience as part of the Georgia Institute of Technology Online Master of Science in Analytics program. I started this program in the fall of 2019, and graduated in December, 2021. My final GPA was 3.91. This document is broken down into sections based on the different classes I completed. With additional sections on the application process, and some general comments about the program in general. For people that are interested in this program, I can personally attest to the fact that the OMSA program is rigorous. Despite being completely online, this is not a program of study that you can just stumble through, at least for me. Granted, I am a non-traditional age student (I am 54 years old.) As part of the program, I took classes with several students that were fresh out of college with degrees in Business, CS, Mathematics, or Statistics. The students that I interacted with spent considerable time watching videos, working on homework and studying for tests. I generally spent anywhere from 6-12 hours on the weekends, and about 1-2 hours per day on course work. Of course, “Your mileage may vary.”
My Background
This section is intended to provide some idea of my background and education. Although this is just a single datapoint, I hope it provides an example of a background that was sufficient for being accepted into the program. Feel free to skip it if you’re not interested. Information about specific classes I took starts after this section.
I was born in the US and am a US Citizen. English is my first language, and I speak some German. I graduated from Arvada West High School in 1986. After high school, I attended Colorado State University for one year and then transferred to the University of Northern Colorado. I graduated from UNC in 1991 (yes, 30 years ago) with a Bachelor of Arts degree. My major was mathematics, and my minor was music, my undergrad GPA was 3.5. I returned to Colorado State in 1992-1993 for one year of graduate school in applied mathematics. My grad school GPA was 3.55.
Prior to the OMSA program, I worked professionally as a computer network engineer, software engineer, DBA, systems engineer and full stack developer in the retail, health insurance, medical device, and transportation and logistics industries. I spent 15 years as an independent contractor in software engineering in Minneapolis, Minnesota.
I applied to the program in the Spring of 2018, and was accepted on August 8, 2018. My first semester was Spring of 2019. Prior to my first semester in the OMSA program, I enrolled in the EdX public course for CSE6040, Computing for Analytics. I essentially audited the CSE6040 fall semester of 2018.
Semester 1 (Spring 2019): CSE6040 Computing for Data Analytics
As mentioned above, I audited CSE6040 in the fall of 2018. The homework assignments for Spring 2019 were all essentially the same as fall 2018, so I had already done them. That made the course significantly easier and less time consuming. The exams, however, were not the same. The midterm and final exams for online “Audited” CSE6040 were much easier than the actual 2019 course. In both cases, the exams were were openbook exams that were electronically graded. We had 24 hours to complete both the midterm and the final. Despite being “open book” those exams were extremely difficult. In many cases, just solving the problem wasn’t enough. We were required to solve the problems and our algorithms had to complete in under a given amount of time. I spent well over 8 hours working on the midterm exam, and over 10 hours on the final. Because the exams were auto-graded, and you could submit your answers as often as you wished, it made the tests somewhat easier, but only slightly so. IMO, the exams were far too difficult. They didn’t really test the content that was covered in the course. I managed to get close to 100% on the midterm and close to 80% on the final. I felt that this course was one of several that were intended to “weed out” weaker students. The TAs in this course were NOT excellent. They were arrogant, and difficult to communicate with. They bordered on being insulting to students. The professor for the course found out that he was going to be teaching it at the last minute, and I don’t feel like he really ever got completely up to speed. Several of the courses that I took had instructors that seem to be only slightly engaged in the course. That is one of the problems with the program. The only notible exceptions were CDA (Dr. Xie), Digital Marketing (Buchannan), and Business Fundamentals (Flurry). In those classes, the instructors were VERY engaged with students.
Topics covered:
(Most topics were done in Python.) Python Review, Pairwise Association, Math/Floating Point Calculation, Web Mining and Regular Expressions, Pandas and SQL, Visualization, Relational Data, NumPy/SciPy, Ranking, Linear Regression, Classification, Clustering, Compression/Dimensionality Reduction, Eigenfaces.
ISYE6501 Introduction to Analytics Modeling
Intro to Modeling was one of the best courses that I took in the program. This course covered topics in machine learning (supervised and unsupervised) time series analysis, PCA (at a high level), basic simulation and optimization. The course focused on R initially, and then later incorporated python. Dr. Sokol’s presentations were outstanding, and the TAs were particularly helpful in this course. I hated R going into this course, and had very little experience with it. I now really like R, and use it extensively particularly for exploratory data analysis.
Interestingly, nearly 80% of what was covered in ISYE 6501 was covered again in other courses like Regression, Simulation, and Computational Data Analytics. Those courses covered the individual topics in much more detail. I found myself in those classes going back to my ISYE 6501 notes. I should also mention that the TAs in this course were excellent.
After I completed the semester, I swore that I would NOT take two courses again at the same time in the program. Despite the fact that audited CSE6040 prior the semester, the amount of work was extreme. I would strongly discourage anyone from taking them together unless you have a great deal of time.
Topics Covered:
(Most topics were done with R.) Classification, Validation, Clustering, Dimensionality Reduction, Time Series (ARIMA), Basic Regression, Advanced Data Prep, Advanced Regression, Tree-based models, Variable Selection, Design of Experiments, Probability based models, Optimization, Specific Use Cases
Semester 2 (Summer 2019): MGT8803 Business Fundamentals for Analytics.
Having wrapped up the two courses in the spring, I next took the business fundamentals course. I didn’t have much academic experience with business courses, but I figured that it would not be too difficult. I was wrong. The Business Fundamentals course was very challenging due to my complete lack of any accounting or other business experience. This course touched on reading balance sheets, understanding where different business accounts would be listed. It also covered topics in accounting like economic value and net-present value. The course touched on difference-in-difference and regression analysis particularly as they related to business topics (for example alpha and beta values of stocks, as well as the capital asset pricing model). The semester ended with a section on business strategy. I just barely made an A in this course. It was FAR more difficult than I expected… that said, I did find it very interesting. This course was a big reason that I decided to focus on the business track in the OMSA program. Contrary to what many people think, some of the business courses in the program are VERY challenging. (In all fairness, some of the business courses are fairly easy too.)
Topics Covered:
Financial Accounting, Managerial Accounting, Financial Analytics Techniques, Entrepreneurial Finance, Business Strategy.
Semester 3 (Fall 2019): ISYE 6644 Simulation.
My third semester in the program was spent taking simulation. Based on the name, I figured this would be a relatively easy course focused on programming. I had read some reviews on OMScentral.com to the contrary. (BTW, omscentral is an invaluable resource for any OMS student!) Despite these warnings, I found myself thinking “how hard can simulation be?” I quickly found out the answer. Simulation is anything but easy. All of the calculus and linear algebra stuff they warn you about, and also the probability theory stuff comes up in this course. If you are not at least reasonably familiar with statistics prior to this course, you should spend time getting up to speed. I was not prepared for the level of stats in this course despite having had probability and statistics in college, and having tutored stats my first time in graduate school. If you don’t know what a gamma distribution is, or the difference between a CDF and a MDF, you are likely not ready for the material in this course, and you will need to work hard to keep up. There is a section of this course that focuses on ARENA (simulation package/program.) That part of the course is relatively easy, and actually quite fun.
Despite being really difficult, Simulation was the most enjoyable course in my OMSA program. Dr. Goldman is absolutely hilarious. And, as he states, there is always a curve. This was the only B I received in the OMSA program… and I worked extremely hard for that B.
Topics Covered:
Whirlwind Tour of Simulation, Bootcamps (Calculus, Integration, Probability, Simulating Random Variables, Expected Values, Functions of Random Variables, Jointly Distributed Random Variables, Conditional Expectations, Covariance and Correlation, Probability Distributions, Limit Theorems, Estimation, Maximum Likelihood, Confidence Intervals, Differential Equations, Monte Carlo Integration, Simulating Pi, Single Queue Servers, Inventory Systems (s,S), Simulating Random Variables, Simulation using Spreadsheets. General Simulation Principles, Arena Simulation Language, Random Variable Generation, Bivariate Random Variable Generation, Input Analysis, Output Analysis, Comparing Systems (statistical testing). (See what I mean. It was tough!)
Semester 4 (Spring 2020): ISYE 6414 Regression Analysis.
There are two kinds of people in the OMSA program, those that really like Dr. Serban, and those that don’t. Personally, I really enjoyed this course. There is a great deal to regression, and Dr. Serban’s course goes much deeper than the ISYE6501 introduction to the topic. I use material that I learned in this course as part of my job. Frankly, I think it should be a required course for the OMSA program. You can read more about this course in OMSCentral, but from my perspective, this was a good course. The workload was reasonable, but not too light. The midterm was VERY challenging, but the final was a bit easier. I think this was a solid course.
Topics Covered:
(Most programming was done in R.) Linear Regression and ANOVA, Multiple Linear Regression (Estimation and Statistical Inference), Generalized Linear Models (Logistic, Poisson, etc.), Variable Selection.
Semester 5 (Summer 2020): MGT 6203 Data Analytics for Business
I have mixed feelings about this course. When I took it, it was the last time they were offering it prior to re-working the content. The course was significantly easier than business fundamentals, but it was sort of all over the place.
Topics Covered:
Linear Regression, Indicator Variables and Interaction Terms, Nonlinear Transformation Models, Logistic Regression, Randomized Controlled Experiments and Natural Experiments, Introduction to Measuring Risk, Factor Investing, Marketing and Advertising, Integrated Digital Marketing and Predictive Marketing, Managing Queues, Forecasting Demand, Statistical Process Control, and Inventory Management. Please note: This course has since been reworked. The topics are now likely very different from those above.
Semester 6 (Fall 2020): ISYE 6740 Computational Data Analytics
Computational data analytics was a very challenging course, and covered a number of models that were touched on in ISYE6501, but in much more depth. Dr. Xie was by far the most responsive professor I had in the OMSA program. She DIRECTLY answered questions, and didn’t completely rely on TAs. That was really great. The course did at least one mathematical proof that was very challenging. The homework assignments were NOT easy, but doable. I was fortunate to have a study partner in this course to bounce ideas off of, and to get clarification.
Actually, a word about study partners. I very much followed the Georgia Tech honor code. I didn’t share code with others, and did my very best to focus on the main purpose of learning the materials in the course. I know the temptation to cheat can be very strong. There are tons of sites on the internet that have code, etc. I decided not to go that route, and I would ask that others please do the same. IMO, the honor code helps to prevent the “dilution” of quality of the program. Yes, the OMSA is hard. But once you have achieved it, you will want to be able to say that you completed the program and have that carry some meaning. So, don’t cheat. It only hurts yourself and others in the long run.
For this course, I also did a project that involved analysis of audio data. I was very proud of this work, and it led to further research in this area that eventually was the topic for a patent that I submitted on behalf of my employer. I am proud of that work, and thankful that my academic research as part of 6740 helped form the basis of my further research and patent application.
Topics Covered:
Clustering and K-Means, Spectral Clustering, PCA and Nonlinear Dimensionality Reduction, Density Estimation, Gaussian Mixture and EM Models, Basics of Optimization Theory, Naïve Bayes Classification, Logistic Regression, Support Vector Machines, Feature Selection, Anomaly Detection, Boosting Algorithms and AdaBoost, Random Forest, Bias-Variance Tradeoff, Special Project.
MGT8823 Continuous Improvement:
I bent my rule of not taking two courses at once in Fall of 2020 because CI had a reputation of being fairly easy. Continuous Improvement is an interesting course. The individual who appeared in the videos for the course (and calls himself a “professor”) no longer teaches at Georgia Tech. This video lecturer didn’t hold a PhD or have tenure, but rather is/was a Six Sigma Master Blackbelt. The actual professor for the course (responding to comments, and grading assignments with his TAs) was OK, but I don’t know if he was entirely prepared for the course. The best part of this course was reading “Moneyball”, and doing reports on it. The worst part of this course, frankly, was the assignments and the videos. I found the instructor to be borderline offensive, though I don’t think that was his intention. I did get some value out of this course, and the six sigma approach. Unfortunately, the instructor (in the videos) didn’t resonate with me. As part of this course, I was awarded a Six Sigma Yellow Belt.
Topics Covered:
Develop and categorize KPIs, Understand Y f(x), Identify Y and X variables, Analyze Y f(x) relationships, Apply the four methods of continuous improvement (DMAIC, Lean, DFSS-DMADV, “GO DO!”), utilize Lean and Six Sigma techniques, apply the DMAIC methodology.
Semester 7 (Spring 2021): CSE 6242 (Data and Visual Analytics)
Oh boy. This course has a reputation for being a killer… and it is well earned. I wish that I could say that I loved this course, but I didn’t. It was just plain hard, and from my experience in graduate school in the past I would have to say that the course felt like a “weed out course.” This course covered a TON of topics that were just all over the place. The project was incredibly difficult because we were not given very clear directions on what was expected. My team actually received an F on the midterm presentation. We were able to plead our case up to a C-. The paper was very difficult to write simply because we had no real clear idea of how to write it. We ended up developing a visualization on Amazon.com product reviews for pet products based on a source of data we found on the web. Our project was a kind of recommendation engine that combined several different ML models (including LDA topic extraction, regression and sentiment analysis to name a few.). I did a tremendous amount of technical work on the data manipulation, training and backend API for the project, and left the front end UI (D3…. If you don’t know what that is, you will) and documentation to others. I ended up getting an A in this course, but it was brutal. If I had taken this course in my initial semester, I don’t know if I would have completed the program.
One thing to note here is that, just when you thought that the difficult part of the course was over, Homework 4 includes implementing both the Page Rank and Random Forest algorithms from scratch. Many students that I know of didn’t complete the Random Forest question, and several didn’t even attempt it. That homework assignment was brutal.
Topics Covered:
Analytics Building Blocks, Data Science Buzzwords, Data Collection, SQLite, Data Cleaning, Data Integration, Visualization (D3), Hadoop, Pig, Hive, Spark, HBase, Classification, Clustering, Graph Methods, Ensemble Methods, Text Analytics, Pagerank, Random Forest (from scratch.)
Semester 8 (Summer 2021): MGT 6311 (Digital Marketing)
I thought that Digital Marketing would be a cakewalk and not be that interesting. It was surprisingly challenging and very interesting. The course covered various topics in digital marketing, and included a number of specific case studies. I found the case studies to be really interesting, and it was so great to study actual current events and business processes. Teaching a course like this is incredibly difficult simply because the topic evolves so quickly. The instructor took the time to try and keep everyone up to speed on current trends, and religiously held office hours to discuss the content and timely topics in digital marketing. The worst, but most valuable part of this course was the fact that I made a mistake when setting up a Google Analytics account to promote my blog, datascience.netlify.com. I ended up spending $300 by mistake. I won’t make that mistake again… but on the bright side, I did learn a valuable lesson.
Topics Covered:
Digital marketing strategy, online brand building and storytelling, social media marketing, online lead generation, mobile marketing, digital thought leadership.
Semester 9 (Fall 2021): MGT 6748 (Business Analytics Practicum)
I just completed this course. It is, as many have said, whatever you put into it. I watched the optional videos (for 20% grade boost.). I would recommend the videos, and not just because of the grade boost. The content (except for the Accenture video) was quite good.
My experience with the practicum really came down to having a good person to work with at my company. My project involved building analytic tools for construction project managers based on cost flow projections. This work was not new research, but built on some prior research. My final report was about 5 pages long, and included a bunch of graphs and diagrams.
The worst part of this course was the time delay between when I completed the course, and when I finally received my grade. This was almost a month of waiting on pins and needles for my final grade. The VAST majority of people that take the practicum get an A. That said, when it is your last grade required for the Masters Degree that is 30 years overdue, you tend to be a bit nervous.
Conclusion
The OMSA program at Georgia Tech provides students with a rigorous foundation in analytics. The program is interdisciplinary and teaches aspects of mathematics, statistics, computer science, and business. The program allows students to focus on analytics tools (quantitative analytics), business, or computational (big data) analytics. The program is designed to be completed in 2 years, but typically takes students with full time jobs significantly more time.
The OMSA program is one of the most affordable Masters programs related to data science, and has earned reputation for high quality. As a result, the program is very competitive to get into, and the quality of the students in the program tends to be very high. Coursework for the program is made up of 15 hours of core coursework, 15 hours of electives, and a 6 credit practicum project.
The students in the program are from diverse backgrounds, countries and cultures. The faculty consists of award winning educators and researchers as well as a large number of teaching assistants. The entire program is offered online. Courses vary, but typically included video lectures, assigned readings, homework assignments (either computer or manually graded) quizzes and online proctored exams. (In proctored exams, students are monitored with webcams and microphones.). Some courses allowed open notes on exams, some allowed limited notes on exams, and some courses were closed-book exams only.
I would encourage anyone looking for an affordable masters degree in analytics to look into the program. It is challenging, but well worth the effort.
Acknoledgements
I would like to thank Mac Johnson, Thom Dietrich, and Adam Selke for writing letters of recommendation on my behalf as part of my application to the program. I would like to thank my classmates and teammates especially Christy, Serena, Mark, Nancy, Kim, Andrew and Laura for their helpful hints and encouragement. I would also like to thank all the professors, instructors, and TAs that helped me learn so much.
Thanks to Jacob Osterman, Thomas Fansler, Jim Coleman and Michael N. Smith for their support on behalf of my employer, Trimble, Inc. Special thanks also to Anne Hunt for her support and encouragement throughout the program and setting an example of getting your masters done while working full time!
I would especially like to thank my kids for tolerating their dad doing homework while on vacation and the weekends, and putting up with my jokes.
Lastly, and most importantly, I would like to thank my wife, Tracy B. Porter, for suggesting that I apply to the program, for her tolerance of me when I was less than at my best, and for always believing in me and supporting me regardless of what degrees I hold. (I love you)^3.