FFORMA

Feature-based forecast model averaging

Research output: Contribution to journalArticleResearchpeer-review

8 Citations (Scopus)

Abstract

We propose an automated method for obtaining weighted forecast combinations using time series features. The proposed approach involves two phases. First, we use a collection of time series to train a meta-model for assigning weights to various possible forecasting methods with the goal of minimizing the average forecasting loss obtained from a weighted forecast combination. The inputs to the meta-model are features that are extracted from each series. Then, in the second phase, we forecast new series using a weighted forecast combination, where the weights are obtained from our previously trained meta-model. Our method outperforms a simple forecast combination, as well as all of the most popular individual methods in the time series forecasting literature. The approach achieved second position in the M4 competition.

Original languageEnglish
Number of pages7
JournalInternational Journal of Forecasting
DOIs
Publication statusAccepted/In press - 2019

Keywords

  • Forecast combination
  • M4 competition
  • Meta-learning
  • Time series features
  • XGBoost

Cite this

@article{48409661e161440a96a10147c2c3cdcc,
title = "FFORMA: Feature-based forecast model averaging",
abstract = "We propose an automated method for obtaining weighted forecast combinations using time series features. The proposed approach involves two phases. First, we use a collection of time series to train a meta-model for assigning weights to various possible forecasting methods with the goal of minimizing the average forecasting loss obtained from a weighted forecast combination. The inputs to the meta-model are features that are extracted from each series. Then, in the second phase, we forecast new series using a weighted forecast combination, where the weights are obtained from our previously trained meta-model. Our method outperforms a simple forecast combination, as well as all of the most popular individual methods in the time series forecasting literature. The approach achieved second position in the M4 competition.",
keywords = "Forecast combination, M4 competition, Meta-learning, Time series features, XGBoost",
author = "Pablo Montero-Manso and George Athanasopoulos and Hyndman, {Rob J.} and Talagala, {Thiyanga S.}",
year = "2019",
doi = "10.1016/j.ijforecast.2019.02.011",
language = "English",
journal = "International Journal of Forecasting",
issn = "0169-2070",
publisher = "Elsevier",

}

FFORMA : Feature-based forecast model averaging. / Montero-Manso, Pablo; Athanasopoulos, George; Hyndman, Rob J.; Talagala, Thiyanga S.

In: International Journal of Forecasting, 2019.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - FFORMA

T2 - Feature-based forecast model averaging

AU - Montero-Manso, Pablo

AU - Athanasopoulos, George

AU - Hyndman, Rob J.

AU - Talagala, Thiyanga S.

PY - 2019

Y1 - 2019

N2 - We propose an automated method for obtaining weighted forecast combinations using time series features. The proposed approach involves two phases. First, we use a collection of time series to train a meta-model for assigning weights to various possible forecasting methods with the goal of minimizing the average forecasting loss obtained from a weighted forecast combination. The inputs to the meta-model are features that are extracted from each series. Then, in the second phase, we forecast new series using a weighted forecast combination, where the weights are obtained from our previously trained meta-model. Our method outperforms a simple forecast combination, as well as all of the most popular individual methods in the time series forecasting literature. The approach achieved second position in the M4 competition.

AB - We propose an automated method for obtaining weighted forecast combinations using time series features. The proposed approach involves two phases. First, we use a collection of time series to train a meta-model for assigning weights to various possible forecasting methods with the goal of minimizing the average forecasting loss obtained from a weighted forecast combination. The inputs to the meta-model are features that are extracted from each series. Then, in the second phase, we forecast new series using a weighted forecast combination, where the weights are obtained from our previously trained meta-model. Our method outperforms a simple forecast combination, as well as all of the most popular individual methods in the time series forecasting literature. The approach achieved second position in the M4 competition.

KW - Forecast combination

KW - M4 competition

KW - Meta-learning

KW - Time series features

KW - XGBoost

UR - http://www.scopus.com/inward/record.url?scp=85066097439&partnerID=8YFLogxK

U2 - 10.1016/j.ijforecast.2019.02.011

DO - 10.1016/j.ijforecast.2019.02.011

M3 - Article

JO - International Journal of Forecasting

JF - International Journal of Forecasting

SN - 0169-2070

ER -