Time series feature learning with labeled and unlabeled data

Haishuai Wang, Qin Zhang, Jia Wu, Shirui Pan, Yixin Chen

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Time series classification has attracted much attention in the last two decades. However, in many real-world applications, the acquisition of sufficient amounts of labeled training data is costly, while unlabeled data is usually easily to be obtained. In this paper, we study the problem of learning discriminative features (segments) from both labeled and unlabeled time series data. The discriminative segments are often referred to as shapelets. We present a new Semi-Supervised Shapelets Learning (SSSL for short) model to efficiently learn shapelets by using both labeled and unlabeled time series data. Briefly, SSSL engages both labeled and unlabeled time series data in an integrated model that considers the least squares regression, the power of the pseudo-labels, shapelets regularization, and spectral analysis. The experimental results on real-world data demonstrate the superiority of our approach over existing methods.

Original languageEnglish
Pages (from-to)55-66
Number of pages12
JournalPattern Recognition
Volume89
DOIs
Publication statusPublished - May 2019

Keywords

  • Classification
  • Feature selection
  • Semi-supervised learning
  • Time series

Cite this

Wang, Haishuai ; Zhang, Qin ; Wu, Jia ; Pan, Shirui ; Chen, Yixin. / Time series feature learning with labeled and unlabeled data. In: Pattern Recognition. 2019 ; Vol. 89. pp. 55-66.
@article{89a4e0972e8b484eb31477465da57deb,
title = "Time series feature learning with labeled and unlabeled data",
abstract = "Time series classification has attracted much attention in the last two decades. However, in many real-world applications, the acquisition of sufficient amounts of labeled training data is costly, while unlabeled data is usually easily to be obtained. In this paper, we study the problem of learning discriminative features (segments) from both labeled and unlabeled time series data. The discriminative segments are often referred to as shapelets. We present a new Semi-Supervised Shapelets Learning (SSSL for short) model to efficiently learn shapelets by using both labeled and unlabeled time series data. Briefly, SSSL engages both labeled and unlabeled time series data in an integrated model that considers the least squares regression, the power of the pseudo-labels, shapelets regularization, and spectral analysis. The experimental results on real-world data demonstrate the superiority of our approach over existing methods.",
keywords = "Classification, Feature selection, Semi-supervised learning, Time series",
author = "Haishuai Wang and Qin Zhang and Jia Wu and Shirui Pan and Yixin Chen",
year = "2019",
month = "5",
doi = "10.1016/j.patcog.2018.12.026",
language = "English",
volume = "89",
pages = "55--66",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier",

}

Time series feature learning with labeled and unlabeled data. / Wang, Haishuai; Zhang, Qin; Wu, Jia; Pan, Shirui; Chen, Yixin.

In: Pattern Recognition, Vol. 89, 05.2019, p. 55-66.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Time series feature learning with labeled and unlabeled data

AU - Wang, Haishuai

AU - Zhang, Qin

AU - Wu, Jia

AU - Pan, Shirui

AU - Chen, Yixin

PY - 2019/5

Y1 - 2019/5

N2 - Time series classification has attracted much attention in the last two decades. However, in many real-world applications, the acquisition of sufficient amounts of labeled training data is costly, while unlabeled data is usually easily to be obtained. In this paper, we study the problem of learning discriminative features (segments) from both labeled and unlabeled time series data. The discriminative segments are often referred to as shapelets. We present a new Semi-Supervised Shapelets Learning (SSSL for short) model to efficiently learn shapelets by using both labeled and unlabeled time series data. Briefly, SSSL engages both labeled and unlabeled time series data in an integrated model that considers the least squares regression, the power of the pseudo-labels, shapelets regularization, and spectral analysis. The experimental results on real-world data demonstrate the superiority of our approach over existing methods.

AB - Time series classification has attracted much attention in the last two decades. However, in many real-world applications, the acquisition of sufficient amounts of labeled training data is costly, while unlabeled data is usually easily to be obtained. In this paper, we study the problem of learning discriminative features (segments) from both labeled and unlabeled time series data. The discriminative segments are often referred to as shapelets. We present a new Semi-Supervised Shapelets Learning (SSSL for short) model to efficiently learn shapelets by using both labeled and unlabeled time series data. Briefly, SSSL engages both labeled and unlabeled time series data in an integrated model that considers the least squares regression, the power of the pseudo-labels, shapelets regularization, and spectral analysis. The experimental results on real-world data demonstrate the superiority of our approach over existing methods.

KW - Classification

KW - Feature selection

KW - Semi-supervised learning

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=85059320212&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2018.12.026

DO - 10.1016/j.patcog.2018.12.026

M3 - Article

VL - 89

SP - 55

EP - 66

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

ER -