Exact and Robust Conformal Inference Methods for Predictive Machine Learning With Dependent Data

TL;DR

This paper extends conformal inference to time series data by incorporating block structures in the permutation scheme, ensuring validity even under serial dependence, thus broadening the applicability of conformal methods.

Contribution

It introduces a randomization-based conformal inference method that accounts for serial dependence in time series, maintaining exact validity under exchangeability and approximate validity otherwise.

Findings

Method retains exact validity for exchangeable data.

Provides approximate validity for dependent time series.

Applicable to a wide range of time series models.

Abstract

We extend conformal inference to general settings that allow for time series data. Our proposal is developed as a randomization method and accounts for potential serial dependence by including block structures in the permutation scheme. As a result, the proposed method retains the exact, model-free validity when the data are i.i.d. or more generally exchangeable, similar to usual conformal inference methods. When exchangeability fails, as is the case for common time series data, the proposed approach is approximately valid under weak assumptions on the conformity score.