Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/336859 
Authors: 
Year of Publication: 
2025
Citation: 
[Journal:] Journal of Forecasting [ISSN:] 1099-131X [Volume:] 45 [Issue:] 1 [Publisher:] Wiley [Place:] Hoboken, NJ [Year:] 2025 [Pages:] 377-390
Publisher: 
Wiley, Hoboken, NJ
Abstract: 
We provide an in‐depth assessment of univariate financial time series analysis via machine learning followed by a thorough discussion beyond the discussion on daily return predictability. We simulate economic time series and present an in‐depth assessment of relevant hyperparameter tuning and study the ability of competing deep learning algorithms to capture econometric properties of financial time series. Also, we assess empirical data and discuss competing approaches in comparison with econometric benchmarks, when the data generating process is unknown. As a result, we assess more than 512,000 in‐sample and out‐of‐sample forecasts for different scenarios of competing network architectures. Drawing on realistic sample sizes, we find that recurrent neural networks with one layer describe a solid alternative to econometric autoregressive moving average (ARMA) approach.
Subjects: 
deep learning
forecasting
machine learning
risk measurement
time series analysis
Persistent Identifier of the first edition: 
Creative Commons License: 
cc-by Logo
Document Type: 
Article
Document Version: 
Published Version

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.