Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP

Erkal Ersoy; Haoyang Li; Mark Edwin Schaffer; Tibor Szendrei

Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP

Erkal Ersoy, Haoyang Li, Mark Edwin Schaffer, Tibor Szendrei

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract

Machine learning methods are being increasingly adopted in economic forecasting. Many learners are available, and a practical issue now presents itself: which one(s) to use? The answer we suggest is ‘stacking regression’ (Wolpert, 1992), an ensemble method for combining predictions of different learners. We show how to use stacking regression in the time series setting. Macroeconomic and financial time series data present their own challenges to forecasting (extreme values, regime changes, etc.), and this presents challenges to
stacking as well. Our findings suggest that using absolute deviations for scoring the base learners performs well compared to stacking on means squared errors. We illustrate this with a Monte Carlo exercise and an empirical application: forecasting US GDP growth around the Covid-19 pandemic.

Original language	English
Title of host publication	Optimal Transport Statistics for Economics and Related Topics
Editors	Ngoc Thach Nguyen, Vladik Kreinovich, Thanh Ha Doan, Duc Trung Nguyen
Publisher	Springer
ISBN (Electronic)	978-3-031-35763-3
ISBN (Print)	978-3-031-35762-6, 978-3-031-35765-7
Publication status	Accepted/In press - 31 Jan 2023

Publication series

Name	Studies in Systems, Decision and Control
Publisher	Springer
ISSN (Print)	2198-4182
ISSN (Electronic)	2198-4190

Keywords

stacking regression
machine learning
robust statistics
forecasting

Embargoed Document

Download pdf
Accepted author manuscript, 1.56 MB
Embargo ends: 14/06/25

Cite this

@inbook{f1e311dc57194e5e81f0c79f2814a1ec,

title = "Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP",

abstract = "Machine learning methods are being increasingly adopted in economic forecasting. Many learners are available, and a practical issue now presents itself: which one(s) to use? The answer we suggest is {\textquoteleft}stacking regression{\textquoteright} (Wolpert, 1992), an ensemble method for combining predictions of different learners. We show how to use stacking regression in the time series setting. Macroeconomic and financial time series data present their own challenges to forecasting (extreme values, regime changes, etc.), and this presents challenges tostacking as well. Our findings suggest that using absolute deviations for scoring the base learners performs well compared to stacking on means squared errors. We illustrate this with a Monte Carlo exercise and an empirical application: forecasting US GDP growth around the Covid-19 pandemic.",

keywords = "stacking regression, machine learning, robust statistics, forecasting",

author = "Erkal Ersoy and Haoyang Li and Schaffer, {Mark Edwin} and Tibor Szendrei",

year = "2023",

month = jan,

day = "31",

language = "English",

isbn = "978-3-031-35762-6",

series = "Studies in Systems, Decision and Control",

publisher = "Springer",

editor = "Nguyen, {Ngoc Thach} and Vladik Kreinovich and Doan, {Thanh Ha} and Nguyen, {Duc Trung}",

booktitle = "Optimal Transport Statistics for Economics and Related Topics",

}

Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP. / Ersoy, Erkal; Li, Haoyang; Schaffer, Mark Edwin et al.
Optimal Transport Statistics for Economics and Related Topics. ed. / Ngoc Thach Nguyen; Vladik Kreinovich; Thanh Ha Doan; Duc Trung Nguyen. Springer, 2023. (Studies in Systems, Decision and Control).

Research output: Chapter in Book/Report/Conference proceeding › Chapter

TY - CHAP

T1 - Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP

AU - Ersoy, Erkal

AU - Li, Haoyang

AU - Schaffer, Mark Edwin

AU - Szendrei, Tibor

PY - 2023/1/31

Y1 - 2023/1/31

N2 - Machine learning methods are being increasingly adopted in economic forecasting. Many learners are available, and a practical issue now presents itself: which one(s) to use? The answer we suggest is ‘stacking regression’ (Wolpert, 1992), an ensemble method for combining predictions of different learners. We show how to use stacking regression in the time series setting. Macroeconomic and financial time series data present their own challenges to forecasting (extreme values, regime changes, etc.), and this presents challenges tostacking as well. Our findings suggest that using absolute deviations for scoring the base learners performs well compared to stacking on means squared errors. We illustrate this with a Monte Carlo exercise and an empirical application: forecasting US GDP growth around the Covid-19 pandemic.

AB - Machine learning methods are being increasingly adopted in economic forecasting. Many learners are available, and a practical issue now presents itself: which one(s) to use? The answer we suggest is ‘stacking regression’ (Wolpert, 1992), an ensemble method for combining predictions of different learners. We show how to use stacking regression in the time series setting. Macroeconomic and financial time series data present their own challenges to forecasting (extreme values, regime changes, etc.), and this presents challenges tostacking as well. Our findings suggest that using absolute deviations for scoring the base learners performs well compared to stacking on means squared errors. We illustrate this with a Monte Carlo exercise and an empirical application: forecasting US GDP growth around the Covid-19 pandemic.

KW - stacking regression

KW - machine learning

KW - robust statistics

KW - forecasting

M3 - Chapter

SN - 978-3-031-35762-6

SN - 978-3-031-35765-7

T3 - Studies in Systems, Decision and Control

BT - Optimal Transport Statistics for Economics and Related Topics

A2 - Nguyen, Ngoc Thach

A2 - Kreinovich, Vladik

A2 - Doan, Thanh Ha

A2 - Nguyen, Duc Trung

PB - Springer

ER -

Stacking Regression for Time-Series, with an Application to Forecasting Quarterly US GDP

Abstract

Publication series

Keywords

Embargoed Document

Fingerprint

Cite this