In the previous chapters, we focused on regression models, and the majority of emphasis was on the linearity assumption. In what appears that the next extension must be non-linear models, we will instead deviate to recursive partitioning techniques, which are a bit more flexible than the non-linear generalization of the models considered in the earlier chapters. Of course, the recursive partitioning techniques, in most cases, may be viewed as non-linear models.
We will first introduce the notion of recursive partitions through a hypothetical dataset. It is apparent that the earlier approach of the linear models changes in an entirely different way with the functioning of the recursive partitions. Recursive partitioning depends upon the type of problem we have at hand. We develop a regression tree for the regression problem when the output is a continuous variable, as in the linear models. If the output is a binary variable, we develop a classification...