Regression Trees

What about regression trees?

There are implementations in SciPy and different R libraries (for example rpart but how do they actually work intuitively compared with decision trees?

In general the implementation is based on the famous CART (classification and regression trees) and work through using recursive partitions to seperate your continuous response. Once the stopping criterion is reached it will use local regression techniques to finally predict your answer. Essentially they created through piece-wise linear models.

In this manner, decision trees can be modified for a continuous response.