Correlation Between Time Series Features and the sMAPE of the Best Methods of M4 Competition

Simple item page
AudienciaPúblico en generales_ES
CoberturaMéxicoes_ES
Fecha de ingreso2026-10-05T16:33:47Z
Fecha de publicación2026-01-01
ResumenThis study analyzes the correlation between time series characteristics and sMAPE forecasting error for the top five methods in the M4 Competition. We use two sets of financial time series with daily and monthly frequencies from the M4 Competition. Additionally, we extracted key characteristics such as trend, curvature, and seasonal strength, among others, for each series. The sMAPE error is calculated using the results provided by the competition. These variables are used in a Random Forest classifier to train ten different prediction models to identify the best forecasting method for each set of time series based on the forecasting error. The results indicate a strong correlation between the time series characteristics and the performance of the top forecasting methods in the M4 Competition, especially in the monthly time series. These series exhibit a remarkable ability to identify the most accurate forecasting method, suggesting that the characteristics of this set of series can be a determining factor in the selection process of the optimal forecasting method.es_ES
Doihttps://doi.org/10.1007/978-3-032-04056-5_16es_ES
URIhttps://riuat.uat.edu.mx/handle/123456789/5567
Idiomaenes_ES
EditorialSpringer Science and Business Media Deutschland GmbHes_ES
RelaciónStudies in Systems, Decision and Controles_ES
URL relacionadohttps://doi.org/10.1007/978-3-032-04056-5_16es_ES
DerechosAcceso restringido / Suscripción (Metadatos de producción científica)es_ES
Licenciahttp://purl.org/coar/access_right/c_16eces_ES
FuenteStudies in Systems, Decision and Control
Palabra claveFinancial time serieses_ES
Palabra claveRandom forestes_ES
Palabra claveTrend and seasonal strengthes_ES
TítuloCorrelation Between Time Series Features and the sMAPE of the Best Methods of M4 Competitiones_ES
TipoCapítulo Libroes_ES
ArbitradoHa sido Arbitradoes_ES
AutorMartínez-Ponce, Mario Alberto
AutorTerán-Villanueva, J. David
AutorSantiago, Alejandro
AutorPonce Flores, Mirna Patricia
AutorAlvarez-Navarro, Eduardo
AutorMartínez-Ponce, Mario Albertoes_ES
AutorTerán-Villanueva, J. Davides_ES
AutorSantiago, Alejandroes_ES
AutorPonce Flores, Mirna Patriciaes_ES
AutorAlvarez-Navarro, Eduardoes_ES
InstituciónUniversidad Autónoma de Tamaulipas
InstituciónUniversidad Autónoma de Tamaulipases_ES
Rango de páginas395-416es_ES
URL relacionadahttps://doi.org/10.1007/978-3-032-04056-5_16
Tipo de artículoIndexado
Tipo de artículoIndexadoes_ES
Volumen628es_ES

Files

Collections

Documentación Cientifica