Conformal prediction for functional Ordinary kriging

TL;DR

This paper introduces a distribution-free conformal prediction method for functional Ordinary Kriging, providing a way to quantify uncertainty in curve predictions at spatial points, validated through simulations and benchmark data.

Contribution

It proposes a novel conformal prediction approach using modulation functions and conformity scores for functional Kriging, addressing the open issue of uncertainty quantification.

Findings

Demonstrates improved uncertainty quantification over standard methods

Validates approach through simulations and benchmark data

Shows advantages in predictive accuracy and reliability

Abstract

Functional Ordinary Kriging is the most widely used method to predict a curve at a given spatial point. However, uncertainty remains an open issue. In this article a distribution-free prediction method based on two different modulation functions and two conformity scores is proposed. Through simulations and benchmark data analyses, we demonstrate the advantages of our approach when compared to standard methods.