mobKriging-package: Recursive Partitioning of Kriging Models

In schiffner/mobKriging: Recursive Partitioning of Kriging Models

Description

This package provides an interface for model-based recursive partitioning (mob from package party) of Kriging models (km from package DiceKriging). The main functions are mobKriging and predictMobKriging. These allow to build segmented Kriging models by computing decision trees where each terminal node corresponds to a fitted Kriging model and making predictions based on those models. Partitioning of Kriging models can be particularly useful if the data contain numerical as well as categorical explanatory variables. Since Kriging normally requires numerical explanatory variables the categorical variables cannot be taken into account directly but can be used as partitioning variables in mobKriging. Package mobKriging also provides different plots, among others diagnostic plots, for fitted segmented Kriging models.

References

Zeileis, A., Hothorn, T. and Kornik, K. (2008), Model-based recursive partitioning. Journal of Computational and Graphical Statistics, 17(2) 492–514.

Roustant, O., Ginsbourger, D. and Deville, Y. (2012), DiceKriging, DiceOptim: Two R packages for the analysis of computer experiments by Kriging-based metamodeling and optimization. Journal of Statistical Software, 51(1), http://www.jstatsoft.org/.

schiffner/mobKriging documentation built on May 29, 2019, 3:39 p.m.