What are the best ways to teach Operations Research students about technology?

Software tools depend on variety of Scenarios for a real industry problem .It depend on the accuracy required, budget for software tools, efficiency of model required. For learning purposes, many Mathematical Programming tools have been providing Acadmic licences like Gurobi etc.Secondly there is huge usecases have been present in web,from where students could learn.The basic objective of the learner is to model real life case the best way possible.Then should move towards Programming tools.Once ,we are well versed with the model,then we are position to take a decision that what type of Tools will suits best with my objectives.

Excel is a good spreadsheet tool. Good commercial solvers, that I think students can access for free for academic purposes, include CPLEX and Gurobi. AMPL is a nice tool because it makes it easy to model mathematical programs and it can connect to a variety of solvers. ProModel is a nice Discrete Event Simulation (DES) tool because it can provide a visualization of your process.

While I do enjoy using algebraic modeling languages due to their succinctness and domain-specific features, students are best served by using solvers and simulators through general purpose languages. They should learn Python and either Java or C#, because they'll likely have to work on teams that use these languages, and using solvers will just be one part of their jobs. They can always learn modeling languages in addition to these, and it will be easier once they're better programmers.

The tools that students learn today are somewhere between completely antiquated and not very useful, with a very rare exception of tools that used by professors who are deeply embedded in the field and are solving real world problems day to day. Even then, they are probably not using the best tools, but the tools they know how to use. For the students to be successful, in my opinion, you could expose them to any tool, but make it clear that it's just for the sake of solving an exercise at hand. Then teach them how these tools are used today, and let the students experiment and explore so they could find the right tools for their work tomorrow. Excel will do 80% of the work for 80% of the people, so why stress so much about the tools?

In my short experience, Gurobi or CPLEX are some of the best IoT tools to understand the analytical thinking of OR beside the Spreadsheet Solvers. But the most important thing is the underlying rigorous mathematical theory, which is sometimes underrated in the name of overwhelming technology. To understand modeling and scientific computing in OR, we need to deep first into the classical paths of Mathematics. For example, an optimization problem in Image Processing, Clustering Analysis or Supply Chain Analytics can provide a lot of information in current technologies, depending on the conceptual example.

I don't recommend teaching with MATLAB because MATLAB is widely used in data science or OR jobs. Well at least it hasn't been used much in any of the teams that I have worked on. I only know of one teammate that has ever used MATLAB for optimization at work.

Gurobi with Python interface is a great option. IBM CPLEX is a great option as well. To solve large scale optimization problems I believe Gurobi is a better other

Python is king here, for now. Java and C# have a solid presence too. While it may be tempting to stop after Python, everyone in the software world should know at least one statically typed, compiled language too. OR-Tools has gained a lot of traction over the past few years -- winning gold medals in the MiniZinc competition doesn't hurt! On the Python side (and I say this as a contributor) there are newer libraries to look at than PuLP. python-mip is one.

If you're considering leaving school and securing a job in industry, it's best to at least have proficiency in Python ??

Ask them if they are able achieve their goals (or Did they ever use goal-programming or achieve numerous goals at once?). Did they ever use SAS or SAS/OR or SAS/STAT for modeling (how to predict in SAS?) and or Simulations/Monte Carlo Simulations.

This is left out of a lot of programs, unfortunately. Students should be able to connect to and query a relational database, read JSON data from a NoSQL data store, and subscribe to event streams. A lot of building models is structuring and aggregating data correctly. A practitioner needs to be able to do things like load a data frame, aggregate and modify it into coefficients for objectives and constraints, and map the output of an optimizer back into a given format.

In a data science / OR jobs in most positions, there aren't people available to create the data needed for a model. You have to do this for yourself and typically all the data isn't easy to get. If you can't get the data yourself you will fail as a data scientist. So just learning Access or SQL or other data tools is a foundation for success to get started. Based on my painful experience in my early years, I wish I had acquired these skills in school rather than learning on the job.

The best way to learn about getting data, in my view, is through projects that need a lot of it. Students can learn about LEFT JOIN and Pivot Table and so forth, but they won't truly understand the meaning of those operations until given a Terabyte of data, and required to find a needle in the haystack. OR instructors should create problems that contain all kinds of data and sources, and encourage students to try and answers some questions. Help them along the way, but don't teach, let them explore and find the right answers.

When building an optimization model, visualization should be a concern from the first lines of code, in terms of the model and the solution it produces. If it is a LP/MIP, you might want to export the model to a readable file format, such as an LP file. This helps making sure each constraint is working as intended. To visualize the solution, it is often best to start by exporting the solution to an excel file and start building from there.

Good visualization is an exercise in storytelling. Choosing the right visualization to express a finding is as important as using a well-designed statistical test. Remember that results need to be communicated to people who aren't quantitative experts, who are stakeholders in product, finance, and other domains.

Presentation is key and is much more valuable than the process, and potentially even the degree to which the solution is right. If you can sell, you will go further than if you are right. Sad, maybe, but true. People don't trust what they don't understand. Figure out not only how to visualize the end results, but also if possible visualize the process. Explain your work to a 5-year old. The tools are not important. Business people would be happy with an Excel/PowerPoint, with ugly figures and fonts. Give it to them! Finance people might dig Power BI, while academic nerds will prefer Matplotlib. Give it to them! I think it's important to learn early on that your solution is not their problem, and whatever you do, do it for your end user.

These articles contain so many MATLAB references that I'm starting to wonder if they're sponsored by Mathworks? MATLAB is not very popular these days, especially outside of academia.

Students early on should decide on whether they can be the best in their field, or not. If the answer is yes, they should excel at becoming the most knowledgeable and experienced OR specialist. Get extremely good at your craft and stay on top of all the new developments. Be unstoppable, and indispensable. If they are not though, they should learn everything else that is not OR. Understand the basics, and then learn sales and marketing, and become a good speaker, learn to be friendly and kind and likable. Figure out how the world works so when it comes time to solving a hard OR problem, you could hire someone from the first group of the actually brilliant people, and the two of you could form a great team.

In hindsight, I viewed my successes as having a big toolbox. Simulation, optimization, statistical analysis, and combinations of various methods. Sometimes I got a problem that looked like an inventory problem but it was really an optimization problem. The tools didn't matter much - few problems I worked couldn't be solved by existing off-the-shelf software. The real advantage was discussing the problem closely with the client, thinking about it for a week or so, then delivering a prototype and a set of examples. When they were amazed that the solution worked as planned, I just said "It's Just Math"!