I am using regression trees in the context of treebagger. My question is what is the model used to make predictions at the nodes and leaves of individual trees? (As opposed to the model by which predictions from various individual trees are combined to give a prediction for the ensemble.)
In particular, when optimising split functions at the nodes, the documentation describes "Choose(ing) a split to minimize the MSE (Mean squared error) of predictions compared to the training data."
However, fully understand this, it is necessary to know how predictions are made in the first place. There is a lot of documentation there, so maybe I've missed the section where the prediction model is described, but if anyone could either describe it, or point me in the direction of documentation where it is described, I'd be grateful.
Ben
Best Answer