A comparative analysis of state-of-the-art-time series forecasting algorithms

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)
24 Downloads (Pure)

Abstract

In recent years many new algorithms have been developed for applications in speech and image processing which can be repurposed for time series prediction. This paper presents a comprehensive comparative analysis of time series forecasting capabilities of eight such state-of-the-art algorithms - namely: Vanilla Long Short-Term Memory(V-LSTM) Gated Recurrent Unit (GRU), Bidirectional LSTM(BD-LSTM), Auto encoder (AE- LSTM), Convolutional Neural Network LSTM(CNN-LSTM), LSTM with convolutional encoder (ConvLSTM), Attention mechanism networks and the Transformer network. Model performances across five different benchmark datasets including fields of interests such as finance, weather and sales are evaluated. Whether direct or iterative prediction methods are optimal for forecasting is investigated. For efficient model optimization, the asynchronous successive halving algorithm (ASHA) is applied in the training folds in a 10 k-fold cross validation framework. Statistical tests are used to comprehensively compare algorithm performances within and across datasets. We show that whilst there are differences between all models, the differences are insignificant for the top performing models which include the Transformer, Attention, V-LSTM, CNN-LSTM and CV-LSTM. However, the transformer model consistently produces the lowest prediction error. We also show that the iterative multistep ahead prediction method is optimal for long range prediction.

Original languageEnglish
Title of host publicationProceedings - 2022 International Conference on Computational Science and Computational Intelligence, CSCI 2022
PublisherIEEE
Pages89-95
Number of pages7
Edition2022
ISBN (Electronic)979-8-3503-2028-2
ISBN (Print)979-8-3503-2029-9
DOIs
Publication statusPublished online - 25 Aug 2023

Publication series

NameProceedings - 2022 International Conference on Computational Science and Computational Intelligence, CSCI 2022

Bibliographical note

Funding Information:
We are grateful for access to the Tier 2 High Performance Computing resources provided by the Northern Ireland High- Performance Computing (NI-HPC) facility funded by the UK Engineering and Physical Sciences Research Council (EPSRC), Grant Nos. EP/T022175/ and EP/W03204X/1. DC is grateful for the UKRI Turing AI Fellowship 2021-2025 funded by the EPSRC (grant number EP/V025724/1). CM is supported by a Department for Economy Northern Ireland funded PhD Scholarship.

Funding Information:
VIII.ACKNOWLEDGEMENT We are grateful for access to the Tier 2 High Performance Computing resources provided by the Northern Ireland High-Performance Computing(NI-HPC) facility funded by the UK Engineering and Physical SciencesResearch Council (EPSRC), Grant Nos. EP/T022175/ and EP/W03204X/1. DC is grateful for the UKRI Turing AI Fellowship 2021-2025 funded by the EPSRC (grant number EP/V025724/1). CM is supported by a Department for Economy Northern Ireland funded PhD Scholarship.

Publisher Copyright:
© 2022 IEEE.

Keywords

  • time series prediction
  • forecasting
  • multi-horizon
  • attention
  • transformer
  • LSTM

Fingerprint

Dive into the research topics of 'A comparative analysis of state-of-the-art-time series forecasting algorithms'. Together they form a unique fingerprint.

Cite this