Decision Tree relevent classification for this task?

Question

I have some data that consists of two-dimensional points and there are two classes, one of which consists of the points whose first and second value are close enough. Would a decision tree be expected to classify this well or should I go with another aproach?

Answer 1

If I understand correctly, you have two predictors x1 and x2, and you want to see if |x1 - x2| is small, ie |x1 - x2| < alpha for some fixed value of alpha.

You don't really need a model for that, do you? I mean, I've just written it down. But to answer the question, no, a decision tree (at least not one of the default algorithms) won't do well. A decision tree creates splits which are parallel to the axes, and here the boundary between the classes is at a 45-degree angle.