Conformal Multistep-Ahead Multivariate Time-Series Forecasting

Filip Schlembach; Evgueni Smirnov; Irena Koprinska

Conformal Multistep-Ahead Multivariate Time-Series Forecasting

Filip Schlembach, Evgueni Smirnov, Irena Koprinska

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

This paper proposes a method for conformal multistep-ahead multivariate time-series forecasting. The method minimizes the coverage loss when the data exchangeability assumption does not properly hold. This is done by weighting residual quantiles while computing prediction intervals. Preliminary experiments on real data demonstrate the method’s utility.

Original language	English
Title of host publication	11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022
Pages	316-318
Number of pages	3
Volume	179
Publication status	Published - 1 Jan 2022
Event	11th Symposium on Conformal and Probabilistic Prediction with Applications - Brighton, United Kingdom Duration: 24 Aug 2022 → 26 Aug 2022 https://cml.rhul.ac.uk/copa2022/

Publication series

Series	Proceedings of Machine Learning Research
ISSN	2640-3498

Conference

Conference	11th Symposium on Conformal and Probabilistic Prediction with Applications
Abbreviated title	COPA 2022
Country/Territory	United Kingdom
City	Brighton
Period	24/08/22 → 26/08/22
Internet address	https://cml.rhul.ac.uk/copa2022/

Keywords

Conformal Prediction
Time Series Forecasting

Cite this

@inproceedings{72e31fb600d44a86b313f5d42c046249,

title = "Conformal Multistep-Ahead Multivariate Time-Series Forecasting",

abstract = "This paper proposes a method for conformal multistep-ahead multivariate time-series forecasting. The method minimizes the coverage loss when the data exchangeability assumption does not properly hold. This is done by weighting residual quantiles while computing prediction intervals. Preliminary experiments on real data demonstrate the method{\textquoteright}s utility.",

keywords = "Conformal Prediction, Time Series Forecasting",

author = "Filip Schlembach and Evgueni Smirnov and Irena Koprinska",

note = "Publisher Copyright: {\textcopyright} 2022 F. Schlembach, E. Smirnov & I. Koprinska.; 11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022 ; Conference date: 24-08-2022 Through 26-08-2022",

year = "2022",

month = jan,

day = "1",

language = "English",

volume = "179",

series = "Proceedings of Machine Learning Research",

publisher = "Proceedings of Machine Learning Research",

pages = "316--318",

booktitle = "11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022",

url = "https://cml.rhul.ac.uk/copa2022/",

}

Schlembach, F , Smirnov, E & Koprinska, I 2022, Conformal Multistep-Ahead Multivariate Time-Series Forecasting. in 11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022. vol. 179, Proceedings of Machine Learning Research, pp. 316-318, 11th Symposium on Conformal and Probabilistic Prediction with Applications, Brighton, United Kingdom, 24/08/22.

Conformal Multistep-Ahead Multivariate Time-Series Forecasting. / Schlembach, Filip ; Smirnov, Evgueni; Koprinska, Irena.
11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022. Vol. 179 2022. p. 316-318 (Proceedings of Machine Learning Research).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

TY - GEN

T1 - Conformal Multistep-Ahead Multivariate Time-Series Forecasting

AU - Schlembach, Filip

AU - Smirnov, Evgueni

AU - Koprinska, Irena

PY - 2022/1/1

Y1 - 2022/1/1

N2 - This paper proposes a method for conformal multistep-ahead multivariate time-series forecasting. The method minimizes the coverage loss when the data exchangeability assumption does not properly hold. This is done by weighting residual quantiles while computing prediction intervals. Preliminary experiments on real data demonstrate the method’s utility.

AB - This paper proposes a method for conformal multistep-ahead multivariate time-series forecasting. The method minimizes the coverage loss when the data exchangeability assumption does not properly hold. This is done by weighting residual quantiles while computing prediction intervals. Preliminary experiments on real data demonstrate the method’s utility.

KW - Conformal Prediction

KW - Time Series Forecasting

M3 - Conference article in proceeding

VL - 179

T3 - Proceedings of Machine Learning Research

SP - 316

EP - 318

BT - 11th Symposium on Conformal and Probabilistic Prediction with Applications, COPA 2022

T2 - 11th Symposium on Conformal and Probabilistic Prediction with Applications

Y2 - 24 August 2022 through 26 August 2022

ER -