Regression trees and forests for non-homogeneous Poisson processes

We propose tree and random forest methods for non-homogeneous Poisson processes. The splitting criterion is derived from a model with a piecewise constant rate function. A simulation study shows that the new method performs well.