Robust online time series prediction with recurrent neural networks

Tian GUO, Zhao XU, Xin YAO, Haifeng CHEN, Karl ABERER, Koichi FUNAYA

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

138 Citations (Scopus)

Abstract

Time series forecasting for streaming data plays an important role in many real applications, ranging from IoT systems, cyber-networks, to industrial systems and healthcare. However the real data is often complicated with anomalies and change points, which can lead the learned models deviating from the underlying patterns of the time series, especially in the context of online learning mode. In this paper we present an adaptive gradient learning method for recurrent neural networks (RNN) to forecast streaming time series in the presence of anomalies and change points. We explore the local features of time series to automatically weight the gradients of the loss of the newly available observations with distributional properties of the data in real time. We perform extensive experimental analysis on both synthetic and real datasets to evaluate the performance of the proposed method.

Original languageEnglish
Title of host publicationProceedings: 3rd IEEE International Conference on Data Science and Advanced Analytics, DSAA 2016
PublisherIEEE
Pages816-825
Number of pages10
ISBN (Electronic)9781509052066
DOIs
Publication statusPublished - 22 Dec 2016
Externally publishedYes
Event3rd IEEE International Conference on Data Science and Advanced Analytics, DSAA 2016 - Montreal, Canada
Duration: 17 Oct 201619 Oct 2016

Publication series

NameInternational Conference on Data Science and Advanced Analytics (DSAA)

Conference

Conference3rd IEEE International Conference on Data Science and Advanced Analytics, DSAA 2016
Country/TerritoryCanada
CityMontreal
Period17/10/1619/10/16

Bibliographical note

Publisher Copyright:
© 2016 IEEE.

Fingerprint

Dive into the research topics of 'Robust online time series prediction with recurrent neural networks'. Together they form a unique fingerprint.

Cite this