O.R. @Home

Catching cat/dog Locomotive

Original photo by Kalden Swart on Unsplash

During my career, I encountered Operations Research (O.R.) tooling in various environments: banking, supply chains, e-commerce, and even fishery industries. In most of the cases, I felt surprised (and sometimes frustrated) that our customers know very little to nothing about the Operations Research discipline, while almost everyone knew what AI, DS, and ML mean. In the best case, people will assume that this is something to do with machine learning (and they will be partially correct), in the worst case they will guess that this is something to do with operations management (also not totally wrong).

Defining Operations Research (O.R.)

One of the reasons for it is that it is pretty tricky to precisely define what Operations Research is. Is it a specific set of algorithms? Is it a number of solutions to specific problems? I will not try to define what is Operations Research anew, but rather will leave a few leads:

• Operations Research and Analytics
• Operations Research and Machine Learning
• What is O.R.?

What can be missing?

In one of the recent community meetings, a participant raised a valid question: why Operations Research achievements have close to no visibility among the non-academic public? O.R. solutions bring a lot of money in numerous industries, but yet this is not something people are familiar with. At that moment we of course look in the direction of ML and AI, the locomotives of our digitalization times (a bit of paradoxical combination, right?), and a reasonable question is what can we do as Operations Research practitioners in order to achieve better visibility of our projects and solutions, learning on successes of our colleagues?

Some features of ML and AI, which we may be possibly lacking in O.R. are methods availability and efficiency. What do I mean here is that in 10 minutes I can checkout git repository code and train a classifier which is able to differentiate cats and dogs. That’s a non-trivial task! I cannot easily explain why some of the pictures belong to one class or another, but this thing does the job! That’s amazing! I can see it working on a very easy, tangible problem which I understand. I can even make it work for myself, e.g. by tagging my own photos.

Why O.R. at home?

With O.R. we construct optimal delivery routes, build effective supply chain schedules, regulate pricing, execute auctions, support big management decisions in top companies. But how O.R. actually can help me? Can I build something as efficient and as available as the aforementioned cat/dog classifier? I guess yes, and this is my main motivation for this series of blog entitled O.R. at Home. I will try to come up with simple everyday decisions, which we as individuals encounter (e.g. at home), and which O.R. methods can effectively support. I would like to also share some of my knowledge in O.R. solutions prototyping, development, and visualization, and I also hope to get valuable feedback from the community.

Thus I put for my self the following goals for the project(s) in this series:

• O.R. at core. That’s easy: should use some of the O.R. tools
• Useful. Projects should be useful (at least for me). There is no better customer to interview, then yourself.
• Visual. Outcomes of the solution should be visualized
• Interactive if possible. I like interactive tools, and thanks to Jupyter environment that’s a pretty achievable goal

Let’s do O.R. at Home!