Minimax Approach for Semivariogram Fitting in Ordinary Kriging

This research paper aims to analyze the minimax approach used in the semivariogram fitting process that forms one stage of the kriging operation performed for interpolation. The conventional method uses the weighted least squares fit for various theoretical functions such as stable, exponential, spherical. However, several recent approaches have been developed using machine learning regression techniques. This research employs the ordinary kriging technique where the proposed minimax approach is expected to increase the accuracy of the interpolation resulted by reducing the error of the final result. Kriging, which is based on the stochastic method, is widely used for spatial values and has been proven to be a better predicting process than deterministic methods. The novel approach to ordinary kriging discussed here, the minimax approach, is able to increase result accuracy based on the experiments performed. Minimax can predict the weights of the semivariogram values better than the weighted least-squares method and performs faster than machine learning approaches.