Meta‐learning how to forecast time series

Features of time series are useful in identifying suitable models for forecasting. We present a general framework, labelled Feature‐based FORecast Model Selection (FFORMS), which selects forecast models based on features calculated from each time series. The FFORMS framework builds a mapping that re...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of forecasting Jg. 42; H. 6; S. 1476 - 1501
Hauptverfasser: Talagala, Thiyanga S., Hyndman, Rob J., Athanasopoulos, George
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Chichester Wiley Periodicals Inc 01.09.2023
Schlagworte:
Agnosticism
algorithm selection problem
black‐box models
Classification
Forecasting
Machine learning
machine learning interpretability
Mapping
random forest
Time
Time series
visualization
ISSN:0277-6693, 1099-131X
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Features of time series are useful in identifying suitable models for forecasting. We present a general framework, labelled Feature‐based FORecast Model Selection (FFORMS), which selects forecast models based on features calculated from each time series. The FFORMS framework builds a mapping that relates the features of a time series to the “best” forecast model using a classification algorithm such as a random forest. The framework is evaluated using time series from the M‐forecasting competitions and is shown to yield forecasts that are almost as accurate as state‐of‐the‐art methods but are much faster to compute. We use model‐agnostic machine learning interpretability methods to explore the results and to study what types of time series are best suited to each forecasting model.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0277-6693
1099-131X
DOI:10.1002/for.2963