HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING
Autor(a) principal: | |
---|---|
Data de Publicação: | 2021 |
Outros Autores: | , , |
Tipo de documento: | Artigo |
Idioma: | eng |
Título da fonte: | Pesquisa operacional (Online) |
Texto Completo: | http://old.scielo.br/scielo.php?script=sci_arttext&pid=S0101-74382021000100209 |
Resumo: | ABSTRACT Recent empirical results show that forecast combinations and cross-learning schemes are winning approaches in the time series field. Although many competition-winning combination methods - with cross-learning or not - use static weights along the forecasting horizon, we could not find extensive work about the effects of using horizon-optimized weights. This paper proposes a forecast combination framework and provides a considerably sizeable empirical investigation into the use of horizon-optimized weights, i.e., weights that may vary over the forecasting horizon. We build on cross-learning, time series clustering and cross-validation to form Horizon-Optimized Convex Combinations (HOC2) of forecasts from five methods: Automated exponential smoothing, Automated ARIMA, Theta, TBATS, and Seasonal naïve. Our combinations were tested with data from the previous M1, M3 and M4 forecast competitions, comprising 104,004 time series with different frequencies and lengths. The results shall be helpful to support future research on how horizon-optimized weights can be used interchangeably with static ones. |
id |
SOBRAPO-1_550fac0432c2e62f89f3557b08f9c707 |
---|---|
oai_identifier_str |
oai:scielo:S0101-74382021000100209 |
network_acronym_str |
SOBRAPO-1 |
network_name_str |
Pesquisa operacional (Online) |
repository_id_str |
|
spelling |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNINGforecast combinationsconvex combinationscross-learningtime series clusteringcross-validationM competitionsABSTRACT Recent empirical results show that forecast combinations and cross-learning schemes are winning approaches in the time series field. Although many competition-winning combination methods - with cross-learning or not - use static weights along the forecasting horizon, we could not find extensive work about the effects of using horizon-optimized weights. This paper proposes a forecast combination framework and provides a considerably sizeable empirical investigation into the use of horizon-optimized weights, i.e., weights that may vary over the forecasting horizon. We build on cross-learning, time series clustering and cross-validation to form Horizon-Optimized Convex Combinations (HOC2) of forecasts from five methods: Automated exponential smoothing, Automated ARIMA, Theta, TBATS, and Seasonal naïve. Our combinations were tested with data from the previous M1, M3 and M4 forecast competitions, comprising 104,004 time series with different frequencies and lengths. The results shall be helpful to support future research on how horizon-optimized weights can be used interchangeably with static ones.Sociedade Brasileira de Pesquisa Operacional2021-01-01info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersiontext/htmlhttp://old.scielo.br/scielo.php?script=sci_arttext&pid=S0101-74382021000100209Pesquisa Operacional v.41 2021reponame:Pesquisa operacional (Online)instname:Sociedade Brasileira de Pesquisa Operacional (SOBRAPO)instacron:SOBRAPO10.1590/0101-7438.2021.041.00245564info:eu-repo/semantics/openAccessValle dos Santos,Rafael de O.Araujo F.,Celso F.Accioly,Ricardo M. S.Oliveira,Fernando Luiz Cyrinoeng2021-11-24T00:00:00Zoai:scielo:S0101-74382021000100209Revistahttp://www.scielo.br/popehttps://old.scielo.br/oai/scielo-oai.php||sobrapo@sobrapo.org.br1678-51420101-7438opendoar:2021-11-24T00:00Pesquisa operacional (Online) - Sociedade Brasileira de Pesquisa Operacional (SOBRAPO)false |
dc.title.none.fl_str_mv |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
title |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
spellingShingle |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING Valle dos Santos,Rafael de O. forecast combinations convex combinations cross-learning time series clustering cross-validation M competitions |
title_short |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
title_full |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
title_fullStr |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
title_full_unstemmed |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
title_sort |
HORIZON-OPTIMIZED WEIGHTS FOR FORECAST COMBINATION WITH CROSS-LEARNING |
author |
Valle dos Santos,Rafael de O. |
author_facet |
Valle dos Santos,Rafael de O. Araujo F.,Celso F. Accioly,Ricardo M. S. Oliveira,Fernando Luiz Cyrino |
author_role |
author |
author2 |
Araujo F.,Celso F. Accioly,Ricardo M. S. Oliveira,Fernando Luiz Cyrino |
author2_role |
author author author |
dc.contributor.author.fl_str_mv |
Valle dos Santos,Rafael de O. Araujo F.,Celso F. Accioly,Ricardo M. S. Oliveira,Fernando Luiz Cyrino |
dc.subject.por.fl_str_mv |
forecast combinations convex combinations cross-learning time series clustering cross-validation M competitions |
topic |
forecast combinations convex combinations cross-learning time series clustering cross-validation M competitions |
description |
ABSTRACT Recent empirical results show that forecast combinations and cross-learning schemes are winning approaches in the time series field. Although many competition-winning combination methods - with cross-learning or not - use static weights along the forecasting horizon, we could not find extensive work about the effects of using horizon-optimized weights. This paper proposes a forecast combination framework and provides a considerably sizeable empirical investigation into the use of horizon-optimized weights, i.e., weights that may vary over the forecasting horizon. We build on cross-learning, time series clustering and cross-validation to form Horizon-Optimized Convex Combinations (HOC2) of forecasts from five methods: Automated exponential smoothing, Automated ARIMA, Theta, TBATS, and Seasonal naïve. Our combinations were tested with data from the previous M1, M3 and M4 forecast competitions, comprising 104,004 time series with different frequencies and lengths. The results shall be helpful to support future research on how horizon-optimized weights can be used interchangeably with static ones. |
publishDate |
2021 |
dc.date.none.fl_str_mv |
2021-01-01 |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/article |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
format |
article |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://old.scielo.br/scielo.php?script=sci_arttext&pid=S0101-74382021000100209 |
url |
http://old.scielo.br/scielo.php?script=sci_arttext&pid=S0101-74382021000100209 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.relation.none.fl_str_mv |
10.1590/0101-7438.2021.041.00245564 |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
text/html |
dc.publisher.none.fl_str_mv |
Sociedade Brasileira de Pesquisa Operacional |
publisher.none.fl_str_mv |
Sociedade Brasileira de Pesquisa Operacional |
dc.source.none.fl_str_mv |
Pesquisa Operacional v.41 2021 reponame:Pesquisa operacional (Online) instname:Sociedade Brasileira de Pesquisa Operacional (SOBRAPO) instacron:SOBRAPO |
instname_str |
Sociedade Brasileira de Pesquisa Operacional (SOBRAPO) |
instacron_str |
SOBRAPO |
institution |
SOBRAPO |
reponame_str |
Pesquisa operacional (Online) |
collection |
Pesquisa operacional (Online) |
repository.name.fl_str_mv |
Pesquisa operacional (Online) - Sociedade Brasileira de Pesquisa Operacional (SOBRAPO) |
repository.mail.fl_str_mv |
||sobrapo@sobrapo.org.br |
_version_ |
1750318018472181760 |