Segmentation trees for stratified data
Tree-growing methods or segmentation trees are in general nonparametric classification techniques. A population of individuals or objects is divided into predefined classes which have to be described by a set of qualitative variables. The problem is to explain or to predict the class to which an individual belongs from the known values of the independent variables for this individual and by looking for logical dependencies between the independent variables and the criterion variable. Tree growing methods are recursive and obtain at each step a binary partition of the underlying population, described by one independent variable, which maximizes the quality gain of prediction or explanation of the criterion variable by the independent variables from one step to the next. This quality of prediction or explanation can be seen as a measure for the information content of the tree with respect to the predefined classes. Frequently the population is not only subdivided into the classes to be investigated, but is simultaneously composed of groups of individuals, which are called strata. Generally, when working with samples from these populations, the observations come either from a stratified random sampling or from a single random sampling scheme. Then it is interesting to investigate how the strata influence the classification rules obtained by a tree-segmentation method. This paper presents a generalized recursive tree-building method that includes strata structures in the tree-building algorithm. In each step it combines the maximization of an information content measure for the criterion variable in a new binary partition of the population with the selection of decisional nodes. Each decisional tree node is composed of a set of strata and a rule for individuals in these strata which will jointly explain the criterion variable. The algorithm is developed for classical data descriptions as well as for probabilistic data descriptions for individuals. As output of the presented algorithm each input stratum or class of individuals is described by a set of symbolic objects which describes different ``segments of objects characterized by the values of their variables as well as by certain percentages. For the entire collection see [Zbl 1039.62501].
- Symbolic object description of strata by segmentation trees
- Far beyond the classical data models: symbolic data analysis
- Sector categorization using gradient boosted trees trained on fundamental firm data
- Relative and Absolute Contributions to Aid Strata Interpretation
- Stratification Before Discriminant Analysis: A Must?
This page was built for publication: Segmentation trees for stratified data
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2756648)
