Improved Median Polish Kriging for Simulation Metamodeling

TL;DR

This paper introduces Improved Median Polish Kriging (IMPK), enhancing prediction accuracy in simulation metamodeling by integrating Biharmonic spline interpolation, demonstrated on coal-ash data with superior results over traditional methods.

Contribution

The paper proposes a novel IMPK method that improves mean function estimation in Median Polish Kriging using Biharmonic spline interpolation.

Findings

IMPK outperforms traditional Median Polish Kriging in prediction accuracy.

Cross validation shows significant improvement with IMPK.

Applied successfully to coal-ash data in two dimensions.

Abstract

In simulation, Median Polish Kriging is a technique used to predict unobserved data points in two-dimensional space. The linear behavior of the traditional Median Polish Kriging in the estimation of the mean function in a high grid makes the interpolation of O(1) which has a low order in the prediction and that leads to a high prediction error. Therefore, an improvement in the estimation of the mean function has been introduced using Biharmonic spline interpolation and the new technique has been called Improved Median Polish Kriging (IMPK). The IMPK has been applied to the standard coal-ash data in two-dimension. The novel method gave much better results according to the cross validation results that were obtained when compared with the traditional Median Polish Kriging.