Blogs

Technical Skills for the Analyst

By Abigail Lindner posted 02-08-2020 23:55

  
I am fairly new to the operations research world. In fact, as of writing I am only halfway through my undergraduate years to earn a B.S. in Mathematics. The OR/MS career came to my attention in the middle of my freshman year when I was exploring mathematics-based career options that weren't actuarial science. Since then, I have tried to learn what I can about the vision of the field and the steps I can take to prepare for entry.

The advice I read online emphasizes gaining technical skills. Given the increasing digitization of society and the reliance of OR/MS on the volumes of data to which technology gives us access, this doesn't come as a surprise. Knowing that you need to develop technical skills, however, is one matter; committing to the task requires a bit more work. For one, the technology boom has brought us a panoply of statistical, mathematical, visualization, and analytical tools, and we can't learn all of them. Furthermore, the tools that data folks herald today may be obsolete five years from now. (Consider the decline in popularity of SPSS.) 

Where is an OR/MS novice to start?

I recently wrote a short article on my personal site on 5 basic but key technical skills that someone should be proficient in using before delving into others. You can find it here. I describe R, Python, SQL, Tableau, and Excel, and provide links to a few online beginner courses for each.

I have gathered this information from personal research on skill set expectations for analysts and from the perspective of an undergraduate student. Given my junior experience, I would be happy to hear thoughts from more experienced OR/MS professionals on this topic.
1 comment
40 views

Related Content

Permalink

Comments

Bruce Hartman
08-15-2024 12:32

Abigail, it's great you are pursuing a career skill with OR/MS.   Actuaries are great, but there are many more uses for math.  

I suggest that two foundational skills for OR/MS are integer programming and nonlinear programming.    Of course linear programming must be thoroughly understood before undertaking these two, because both of them refer back to LP as a baseline.  

The concepts you learn in an IP course and an NLP course are the bedrock of modern optimization analysis.  

I'm sure there are excellent references today for these areas that I am not aware of. But for NLP I don't think that Luenberger's text 'Linear and Nonlinear Programming' (2d ed Addison-Wesley 1989) can be topped!   I find it relevant today. The exercises are challenging and require math skills.  When I took NLP in grad school, our use of the book as a text was coupled with programming assignments in C to implement some of the concepts.   I learned a lot that was useful throughout my career.  The details and the math background they depend on are clearly explained. You can't learn that stuff from playing with a package or a programming platform, and it's critical for interpretation. 

The same can be said for a good course in IP and MIP (mixed integer programming, where coefficients and solutions may or may not be required to be integer); the math behind the results is critical to understand.  Unfortunately, the books I used in grad school weren't ones I would refer anyone to.  Especially, I found the classic by Nemhauser and Wolsey 'Integer and Combinatorial Optimization' quite opaque.  

Everything starts with LP, linear programming.  Probably the best book from my era is Chvatal, 'Linear Programming' (Freeman, 1980).  I used Bazaraa Jarvis and Sherali (Wiley, 1990) 'Linear Programming and Network Flows', and it's especially useful for the discussion of networks, which are key in modern analytics.  I think Magnanti's text on networks (Network Flows: Theory, Algorithms, and Applications (Prentice Hall, 1993)) is possibly the very best from my era.

I think you have to learn the math behind these three areas and a few algorithms that result from it.  I find you keep using it and referring to it on every project.  And I can't understand what Tableau or any other package is really doing without the background, much less the new AI algorithms like neural networks.

As for Tableau, it's great, and popular in industry, I hear.  I would consider also Microsoft's PowerBI, which is often free for students if their university uses the MS suite for students.  It's powerful though complicated, but there are lots of free tutorials provided by MS to learn from.