Robust optimization of forecast combinations

Gerrit Tjeerd (Thierry) Post, Selçuk Karabatı, Stelios Arvanitis

Research output: Contribution to journalArticle

Abstract

We develop a methodology for constructing robust combinations of time series forecast models which improve upon a given benchmark specification for all symmetric and convex loss functions. Under standard regularity conditions, the optimal forecast combination asymptotically almost surely dominates the benchmark, and also optimizes the chosen goal function. The optimum in a given sample can be found by solving a convex optimization problem. An application to the forecasting of changes in the S&P 500 volatility index shows that robust optimized combinations improve significantly upon the out-of-sample forecasting accuracy of both simple averaging and unrestricted optimization.

Original languageEnglish
Pages (from-to)910-926
Number of pages17
JournalInternational Journal of Forecasting
Volume35
Issue number3
DOIs
Publication statusPublished - Jul 1 2019

Fingerprint

Robust optimization
Benchmark
Combination of forecasts
Regularity
Out-of-sample forecasting
Optimization problem
Loss function
Volatility index
Forecast combination
Methodology
Forecasting accuracy

Keywords

  • Asymptotic theory
  • Convex optimization
  • Forecast combinations
  • Stochastic dominance
  • Time series analysis
  • Volatility index forecasting

ASJC Scopus subject areas

  • Business and International Management

Cite this

Robust optimization of forecast combinations. / Post, Gerrit Tjeerd (Thierry); Karabatı, Selçuk; Arvanitis, Stelios.

In: International Journal of Forecasting, Vol. 35, No. 3, 01.07.2019, p. 910-926.

Research output: Contribution to journalArticle

Post, Gerrit Tjeerd (Thierry) ; Karabatı, Selçuk ; Arvanitis, Stelios. / Robust optimization of forecast combinations. In: International Journal of Forecasting. 2019 ; Vol. 35, No. 3. pp. 910-926.
@article{8a3d0ad6de714e949d9c3f1475a64691,
title = "Robust optimization of forecast combinations",
abstract = "We develop a methodology for constructing robust combinations of time series forecast models which improve upon a given benchmark specification for all symmetric and convex loss functions. Under standard regularity conditions, the optimal forecast combination asymptotically almost surely dominates the benchmark, and also optimizes the chosen goal function. The optimum in a given sample can be found by solving a convex optimization problem. An application to the forecasting of changes in the S&P 500 volatility index shows that robust optimized combinations improve significantly upon the out-of-sample forecasting accuracy of both simple averaging and unrestricted optimization.",
keywords = "Asymptotic theory, Convex optimization, Forecast combinations, Stochastic dominance, Time series analysis, Volatility index forecasting",
author = "Post, {Gerrit Tjeerd (Thierry)} and Sel{\cc}uk Karabatı and Stelios Arvanitis",
year = "2019",
month = "7",
day = "1",
doi = "10.1016/j.ijforecast.2019.01.007",
language = "English",
volume = "35",
pages = "910--926",
journal = "International Journal of Forecasting",
issn = "0169-2070",
publisher = "Elsevier",
number = "3",

}

TY - JOUR

T1 - Robust optimization of forecast combinations

AU - Post, Gerrit Tjeerd (Thierry)

AU - Karabatı, Selçuk

AU - Arvanitis, Stelios

PY - 2019/7/1

Y1 - 2019/7/1

N2 - We develop a methodology for constructing robust combinations of time series forecast models which improve upon a given benchmark specification for all symmetric and convex loss functions. Under standard regularity conditions, the optimal forecast combination asymptotically almost surely dominates the benchmark, and also optimizes the chosen goal function. The optimum in a given sample can be found by solving a convex optimization problem. An application to the forecasting of changes in the S&P 500 volatility index shows that robust optimized combinations improve significantly upon the out-of-sample forecasting accuracy of both simple averaging and unrestricted optimization.

AB - We develop a methodology for constructing robust combinations of time series forecast models which improve upon a given benchmark specification for all symmetric and convex loss functions. Under standard regularity conditions, the optimal forecast combination asymptotically almost surely dominates the benchmark, and also optimizes the chosen goal function. The optimum in a given sample can be found by solving a convex optimization problem. An application to the forecasting of changes in the S&P 500 volatility index shows that robust optimized combinations improve significantly upon the out-of-sample forecasting accuracy of both simple averaging and unrestricted optimization.

KW - Asymptotic theory

KW - Convex optimization

KW - Forecast combinations

KW - Stochastic dominance

KW - Time series analysis

KW - Volatility index forecasting

UR - http://www.scopus.com/inward/record.url?scp=85065909464&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065909464&partnerID=8YFLogxK

U2 - 10.1016/j.ijforecast.2019.01.007

DO - 10.1016/j.ijforecast.2019.01.007

M3 - Article

VL - 35

SP - 910

EP - 926

JO - International Journal of Forecasting

JF - International Journal of Forecasting

SN - 0169-2070

IS - 3

ER -