Spatial CART classification trees

We propose to extend CART for bivariate marked point processes provide a segmentation of the space into homogeneous areas interaction between marks. While usual tree considers marginal distribution response variable at each node, proposed algorithm, SpatCART, takes account spatial location observations in splitting criterion. introduce dissimilarity index based on Ripley’s intertype K-function quantifying two populations. This used growing step strategy, leads heterogeneity function consistent with original algorithm. Therefore new variant is way explore data as process using binary classification trees. The procedure implemented an R package, and illustrated simulated examples. SpatCART finally applied tropical forest example.