MEF BDA 503 Fall 2023
Fall 2023 semester is just behind us now. I gave BDA 503 for the seventh year in a row (the program itself is 8 years old). The course, this year, just like the previous years, rhymed but did not repeat itself.
Last year I started writing course-end reviews in its own blog; EDA Journal. So, here, I will briefly explain what we did during Fall 2023 semester, what we did new and what we did the same.
What is MEF BDA 503?
MEF University has a Big Data Analytics Masters Program aimed to teach “data science” to professionals from diverse backgrounds. BDA 503 is the code of the must course “Data Analytics Essentials”. See detailed explanation here.
What did we do?
Datasets we used in assignments are
- Each student worked on a data set they proposed at the beginning of the semester.
- We used Global Dietary Database as a new data set.
- Projects: This semester we had 5 groups. Each group selected a data set from a set of alternatives at the first weeks of the semester.
Group Projects
- European Union Candidates — Turkiye and Ukraine: A Comparative Analysis by group Astral Projection
- A Turkey Analysis: Trade Balance & Partners based on World Integrated Trade Systems (WITS) Database by group Yellow Submarine
- PISA 2022: Analysis of OECD Countries by group Sun Forest
What did we do new?
- Instead of a single data set for assignments, we let students choose their own data sets. This way they would be working on data which they are more interested in.
- Instead of a single source of data, a selection from various sources are given to students as their group project data set.
- A new item added to student course feedback survey. I asked a question whether they liked the guest lecturers or not. Because they take a sizable time from the lectures. Apparently it is a time well-spent.
What did we do the same?
- The backbone of the course process stayed largely the same. In order; R, Quarto, dplyr, ggplot2, shiny progress stayed the same. In the last few years I have been pondering on whether to switch to Python (very seriously). Ultimately, I decided that; even though Python is the eventual destination, R is still the best way forward within a short period of time, also bearing in mind that most of my students are not developers.
- Course tools are now pretty much mature. Slack, Github, Github Desktop, Github Pages, Datacamp for Classrooms are all liked and used by the students.
- Guest lectures are here to stay. They said they benefited from the expertise and experience of those who do practical work at their own domains.
- And of course the Course Survey. It is the gauge of the course’s performance.
Conclusion
I think this course matured well enough to be a fundamental data analysis course. Fundamentals (reproducible analysis, clear communication, and future directions) are covered well. It is a good base for advanced topics such as ML, AI, RL and LLMs. Of course it can still be improved but marginal benefits are getting thinner each year given the time allocated and its purpose.
One thing I noticed is that the course is scalable in both ways. I designed it to scale-up (>50 students) so that marginal work is minimal. But I also noticed that it could scale-down (<15) as well. The course has an open code policy through Github repositories. Therefore, any student getting stuck on some error etc. can learn from their peers immediately. This kind of collaboration increases the learning speed dramatically.
What I would wish for the future is to somehow pack this course into an at-will learning module for cohorts with size from 10 to 100. This course can be specialized in a number of domains as well.
