A good friend of mine is a teacher in elementary school and every year she faces the same problem: having to divide her students into 2 classes. She wants to divide them as evenly as possible based on multiple criteria, like:

  • gender
  • scores on math and spelling
  • whether they need extra attention
  • whether they have a learning (dis)advantage
  • whether they have difficult behavior

On top of that, each student can name 3 other students with whom they’d like to be in the same class (and at least 1 of those preferences should be satisfied). Furthermore, the teacher can decide that particular students should or should not be in the same class together.

Obviously, satisfying those conditions while balancing the other criteria, can prove to be a difficult task for a regular human being. Fortunately, there’s computers.

I initially thought this would be perfect for a Genetic Algorithm to solve. I’ve used GAs in the past to optimize electronic circuits so I’m quite familiar with them (and to be honest, I’m not very familiar with any other optimization algorithms), so I decided to give it a go.

However, it quickly turned out that brute-force optimizing those hard criteria for which students should be in the same class as others, doesn’t really work. So I had to come up with some way to satisfy those conditions as an additional step during the optimization.

The optimization bit, I put in a separate package called optime, available on GitHub and PyPI. The rest is in this package, called classevy, available on GitHub. The name of the package refers to my friend. I hope she can try it for next school year and I hope her students won’t mind being guinea pigs in my Python projects.

I invite everyone to try out some examples (there’s a Jupyter Notebook to get started) and let me know if it works and what could be improved. And if you know teachers with the same, or a similar problem, tell them to contact me. I’d be very interested to make the software more general, if there is a need.