> I actually wrote most of the machine learning content at Dataquest (where I work) and I started with k-nearest neighbors because it's way more approachable T

> I actually wrote most of the machine learning content at Dataquest (where I work) and I started with k-nearest neighbors because it's way more approachable That's a great idea. It's actually what I did in an earlier version of the USF course. It worked great. Especially because a decision tree is basically just KNN with a different distance measure (loosely defined) so it can flow well However it turned out that jumping straight to decision trees worked out well too, so I'm happy with the change.