A Comparative Study of Autoregressive and Neural Network Models: Forecasting the GARCH Process

Simple item page
dc.contributor.authorKamalov, Firuz
dc.contributor.authorGurrib, Ikhlaas
dc.contributor.authorMoussa, Sherif
dc.contributor.authorNazir, Amril
dc.date.accessioned2022-10-04T15:05:06Z
dc.date.available2022-10-04T15:05:06Z
dc.date.copyright© 2022
dc.date.issued2022
dc.description.abstractThe Covid-19 pandemic has highlighted the importance of forecasting in managing public health. The two of the most commonly used approaches for time series forecasting methods are autoregressive (AR) and deep learning models (DL). While there exist a number of studies comparing the performance of AR and DL models in specific domains, there is no work that analyzes the two approaches in the general context of theoretically simulated time series. To fill the gap in the literature, we conduct an empirical study using different configurations of generalized autoregressive conditionally heteroskedastic (GARCH) time series. The results show that DL models can achieve a significant degree of accuracy in fitting and forecasting AR-GARCH time series. In particular, DL models outperform the AR-based models over a range of parameter values. However, the results are not consistent and depend on a number of factors including the DL architecture, AR-GARCH configuration, and parameter values. The study demonstrates that DL models can be an effective alternative to AR-based models in time series forecasting. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
dc.identifier.citationKamalov, F., Gurrib, I., Moussa, S., & Nazir, A. (2022). A comparative study of autoregressive and neural network models: Forecasting the GARCH process. Lecture Notes in Computer Science, 13395, 589 - 603. https://doi.org/10.1007/978-3-031-13832-4_48
dc.identifier.isbn978-303113831-7
dc.identifier.issn03029743
dc.identifier.urihttps://doi.org/10.1007/978-3-031-13832-4_48
dc.identifier.urihttp://hdl.handle.net/20.500.12519/708
dc.language.isoen_US
dc.publisherSpringer Science and Business Media Deutschland GmbH
dc.relationAuthors Affiliations : Kamalov, F., Canadian University Dubai, Dubai, United Arab Emirates; Gurrib, I., Canadian University Dubai, Dubai, United Arab Emirates; Moussa, S., Canadian University Dubai, Dubai, United Arab Emirates; Nazir, A., Zayed University, Abu Dhabi, United Arab Emirates
dc.relation.ispartofseriesLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Volume 13395
dc.rightsLicense to reuse the abstract has been secured from Springer Nature and Copyright Clearance Center.
dc.rights.holderCopyright: © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
dc.subjectARIMA
dc.subjectDeep learning
dc.subjectGARCH
dc.subjectNeural networks
dc.subjectTime series forecasting
dc.titleA Comparative Study of Autoregressive and Neural Network Models: Forecasting the GARCH Process
dc.typeConference Paper

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Access Instruction 708.pdf
Size:
44.04 KB
Format:
Adobe Portable Document Format
Description:
Download

Collections

Department of Electrical Engineering