Making Decision Optimization Simple

Optimization (Prescriptive analytics, Operations Research, Decision Optimization) is doing more with less. Which is why most of the time, this makes sense. In a few posts I tried to show the way:

• Low barrier to entry to Optimization
• Should we use a dedicated modeling language or a general purpose language ?

Then I gave some tricks and examples but nothing happened, well not much. Then came 2018 and I tried a very simple story.

This got some impact and got translated into 23 languages (English included) because this was very simple. In Freakonomics, Nobel Prize 2017 Richard Thaler said : "If you want to get people to do something, make it easy. Remove the barriers."

OPL makes decision optimization very simple and let me share very simple examples with those buses and that zoo.

• basic zoo example
• get solution with MP and CP
• data files
• several dat files
• arrays and sets
• tuple sets and asserts
• many tuple set functions
• objects in scripting
• settings : options 1,2,3

• get cplex info after solve
• get the gap
• forall and slicing
• Excel spreadsheets

• CSV files

• CPO and Constraint Programming
• CPO with decimal decision variables
• Scheduling

• Pre and post processing
• Flow control and incremental changes
• Get and set coef
• Flow control and changes with regenerate
• Multi scenarii in .dat

• range constraint
• Logical constraints
• max constraint
• absolute value
• Disjunctions (or)
• If then (and if static condition)
• Quadratic
• all different
• all and any
• lazy constraint
• Modulo constraints in cplex and cpo
• Decision variable in a domain in cplex and cpo
• Save solution in a file
• Conflicts and relaxations

• Hard and soft constraints
• Relax integrity constraints and dual value
• Relax one integer decision variable
• Solvefixed , dual value and reduced cost
• Simulation and what if analysis

• Piecewise linear function
• Non linear functions
• Table function

• Monte Carlo optimization with regards to Monte Carlo Simulation
• Stochastic optimization
• Robust optimization
• Get several solutions
• Enumerate without solution pools
• Get several solutions with Constraint Programming
• Mixing modeling and scripting
• Write and read MST files : warm start
• Warm start through API : addMIPStart
• Warm start with Constraint Programming through API
• Fixed start with MP and Constraint Programming
• Check feasibility with fixed start or with assert
• Adding rows and columns incrementally
• Plain Multiobjective or with settings and then with Constraint Programming
• Semiinteger and semicontinuous
• count true constraints
• count unique values MP and Constraint Programming
• set branch priorities MIP and CP
• array variable indexer size - 3 ways : union , tuple set, decision expression
• get solutions one by one through main or postprocess all feasible
• export model in flow control or without

What's next ? Once you wrote your model in OPL you may run it from the CPLEX Integrated Developemnt Environment (IDE) and the model will either run on your machine, on a CPLEX server or in the CPLEX cloud.

You can also run in command line (oplrun)

And also use oplrun like command line to call OPL CPLEX model in the WML Cloud

From OPL you may call external programs like Python code.

You may also call your OPL model from C++, Java and .NET through rich APIs that let you see OPL objects as native C++, Java and .NET Objects.

You may call your model in the cloud through all APIs : Java, Node.js, Python, R, Rest

You may do external calls from R, Matlab and so on as explained in the tricks. Or use doopl or ticdat to call OPL from Python. Or directly call OPL from R.

You may also call OPL from Planning Analytics or SPSS Modeler or within Decision Optimization Centre

Let's not forget Cloud Pak for Data and Watson Studio. See example.

And OPL is easy but you have many other APIs if you do not want to use OPL!