Multi Split Conformal Prediction

TL;DR

This paper introduces multi split conformal prediction, an approach that aggregates multiple conformal prediction intervals across various data splits to improve reliability and reduce dependence on a single split.

Contribution

It proposes a novel aggregation method for conformal prediction intervals using Markov's inequality, enhancing distribution-free predictive inference in regression.

Findings

Reduces dependence on a single data split.

Provides more reliable predictive intervals.

Maintains computational efficiency.

Abstract

Split conformal prediction is a computationally efficient method for performing distribution-free predictive inference in regression. It involves, however, a one-time random split of the data, and the result depends on the particular split. To address this problem, we propose multi split conformal prediction, a simple method based on Markov's inequality to aggregate single split conformal prediction intervals across multiple splits.