Instructor Q&A: Keith McCormick, Predictive Analytics

Why did you decide to become an instructor?

I guess I've always been involved with and enjoyed teaching. Even in high school and college I did quite a bit of tutoring. Immediately after obtaining my degree, I ran a small test prep business with nearly a dozen employees, one of whom was a sport psychologist who lectured on test anxiety.

I also nearly pursued an academic career, but in the late nineties there were no advanced degree programs in Data Science, the career that I stumbled into. A couple of years ago, UCI needed a highly customized course using software that I was expert in. The project went very well so I started teaching public courses soon after.

What's unique about your teaching style?

Despite decades of teaching experience, including thousands of hours of software instruction, I really consider myself a consultant at heart. The vast majority of my business is still dependent on producing successful outcomes for my clients.

My classes are unique in that I teach the same way that I would when training a new hire. I know that most of my students in the predictive analytics program are not taking my class because they want to pursue a PhD. They are trying to make themselves more valuable to their employers. I assume that they will be doing this in their career so assignments are challenging, but as real-world as I can make them.

However, I think conceptual knowledge is critical so we do a fair amount of reading. I think this surprises students who think data science is only about the math or about following “recipes” for performing data analytics.

What's your favorite lesson to teach and why?

I am currently teaching courses on the Deployment and Data Understanding phases of the Cross-Industry Standard Process for Data Mining (CRISP-DM). I also teach the Introduction to Predictive Analytics course.

My favorite lesson is the opening week of the Data Understanding course, probably because it is my newest course and I spent a lot of time designing it. I love when students discover that seemingly primitive analyses, when performed correctly, can uncover the strangest things about the data.

I also appreciate when they understand what information you must share with a Subject Matter Expert (SME) before you can safely build a model. It's taken me more than 25 years to get there, but I can find weird quirks in a dataset in less than an hour that my clients didn't know were there.

This is not a kind of performance art to impress the client, but rather an uncovering of critical issues that might endanger a project. It is metaphorically like a home inspection before buying a house. I like revealing this new world to groups of potential future colleagues.

What do you find most rewarding about being an instructor?

I don't know if folks will even believe my answer — it is grading. The reason is that although I share Skype calls with quite a few students I don't “meet” all of them one on one. So grading is my primary correspondence with students, especially those that get full points on most assignments since they are less likely to arrange a help session with me.

Therefore, I don't comment only on poor submissions as constructive criticism. I also comment on submissions with perfect scores and explain why the work is good. I even love it when a student with a perfect score on an assignment submits a second version just to clarify something in our correspondence or tries an alternate approach.

When there is an interesting task, but it is perhaps a bit too challenging to make it a course requirement, I offer it as extra credit to encourage this behavior even more. I've also had students turn awful submissions into excellent ones through our correspondence. I emphasize the creative and subjective aspects of predictive analytics.

For those of us that make our living doing this work, the creative aspect dominates. So, when I start to see students turning in professional work, especially when I can detect true improvement from week to week, I find it very rewarding.