Georgia Tech ISYE 6501: Intro to Analytics Modeling – Course Review
Data analytics is a fascinating field. It’s our way of augmenting human intelligence and gaining insights from large amounts of data available in this digital age. It is commonly said that data is the new oil, but data needs to be refined before it can be used. That is what the field of analytics is about i.e., how to refine the data and gain actionable insights from it.
Most STEM professionals have a general understanding of what analytical models and applications are capable of. But I wanted to gain a more comprehensive understanding of the tool-set and the considerations that go into developing and utilizing analytics models.
After some research, I decided to go for ISYE 6501 Intro to Analytics course which is part of the Georgia Tech Master of Science in Analytics program. I recently finished the course, and it has been a very rewarding experience.
The course sets out the following two learning goals:
- Given a business (or other) question, select an appropriate analytics model to answer it, specify the data you will need to solve it, and understand what the model’s solution will and will not provide as an answer
- Given someone else’s use of analytics to address a specific business (or other) question, evaluate whether they have used an appropriate model (and appropriate data) and whether their conclusion is reasonable
A comprehensive overview of the course and a link to the syllabus can be found on the course overview page. Some of the topics that I found especially interesting were:
- Support Vector Machines and K-Nearest Neighbor classification
- Clustering and K-Means algorithm
- Data preparation including data imputation and dealing with heteroscedasticity
- Model validation and cross-validation
- Triple Exponential Smoothing model for time series data forecasting
- Principal Component Analysis
- Variable selection for regression models (LASSO, Elastic Net, Ridge regression)
- CART and Random Forests
- Confusion Matrices
- Design of Experiments
- Using various probability distributions (e.g., Poisson and Weibull distributions)
- Markov Chains
- Optimization and simulation
- Neural Networks and Deep Learning
- Competitive models and basics of Game Theory
The course does require a reasonable amount of consistency and effort. The homework assignments are in R. Most of my experience has been with C# and Python so it did take a little while to get used to R’s syntax and way of doing things. Keep in mind that if you are planning to take this course with a demanding job, you are committing your Saturdays for ~3-4 months. The threshold to pass the edX.org version is fairly easy to do (~60%). But if eventually you’d like to apply it towards degree credit, then you should aim for ~90%. Which means doing all the homework assignments and giving the exams with a good bit of preparation. That is the path that I went for.
Overall, quite impressed by the quality of the course. I highly recommend it.