How a tried and tested solution can yield excellent leads to tackling a day-to-day ML problem

With a lot attention on generative AI and vast neural networks, it is simple to overlook the tried and tested Machine Learning algorithms of yore (they’re actually not that old…). I’d go to this point as to argue that for many business cases, a simple Machine Learning solution will go further than most complex AI implementation. Not only do ML algorithms scale extremely well, the far lower model complexity is what (in my view) makes them superior in most scenarios. Not to say, I even have also had a far easier time tracking the performance of such ML solutions.

In this text, we’ll tackle a classic ML problem using a classic ML solution. More specifically, I’ll show how one can (in just a number of lines of code) discover feature importance inside a dataset using a Random Forest classifier. I’ll start by demonstrating the effectiveness of this system. I’ll then apply a ‘back-to-basics’ approach to indicate how this method works under the hood by making a Decision Tree and a Random Forest from scratch whilst benchmarking the models along the best way.