Ensemble Based Approach for Time Series Classification in Metabolomics

Michael Netzer, Friedrich Hanser, Marc Breit, K.M. Weinberger, Christian Baumgartner, Daniel Baumgarten

Research output: Chapter in Book/Report/Conference proceedingChapterResearchpeer-review

Abstract

BACKGROUND: Machine learning is one important application in the area of health informatics, however classification methods for longitudinal data are still rare. OBJECTIVES: The aim of this work is to analyze and classify differences in metabolite time series data between groups of individuals regarding their athletic activity. METHODS: We propose a new ensemble-based 2-tier approach to classify metabolite time series data. The first tier uses polynomial fitting to generate a class prediction for each metabolite. An induced classifier (k-nearest-neighbor or naïve bayes) combines the results to produce a final prediction. Metabolite levels of 47 individuals undergoing a cycle ergometry test were measured using mass spectrometry. RESULTS: In accordance with our previous work the statistical results indicate strong changes over time. We found only small but systematic differences between the groups. However, our proposed stacking approach obtained a mean accuracy of 78% using 10-fold cross-validation. CONCLUSION: Our proposed classification approach allows a considerable classification performance for time series data with small differences between the groups.

Original languageEnglish
Title of host publicationdHealth 2019 - From eHealth to dHelath
Subtitle of host publicationProceedings of the 13th Health Informatics Meets Digital Health Conference
EditorsDieter Hayn, Alphons Eggerth, Günter Schreier
PublisherIOS Press
Pages89-96
Number of pages8
ISBN (Print)978-1-61499-974-4
DOIs
Publication statusPublished - 2019
EventdHealth 2019 - Schönbrunn Palace, Wien, Austria
Duration: 28 Mar 201929 Mar 2019
http://www.ehealth2017.at

Publication series

NameStudies in health technology and informatics
Volume260

Conference

ConferencedHealth 2019
CountryAustria
CityWien
Period28/03/1929/03/19
Internet address

Fingerprint

Metabolomics
Metabolites
Time series
Ergometry
Informatics
Sports
Mass spectrometry
Learning systems
Mass Spectrometry
Classifiers
Health
Polynomials

Keywords

  • biomarkers
  • classification
  • kinetics
  • time series

ASJC Scopus subject areas

  • Health Information Management
  • Health Informatics
  • Biomedical Engineering

Fields of Expertise

  • Human- & Biotechnology

Cite this

Netzer, M., Hanser, F., Breit, M., Weinberger, K. M., Baumgartner, C., & Baumgarten, D. (2019). Ensemble Based Approach for Time Series Classification in Metabolomics. In D. Hayn, A. Eggerth, & G. Schreier (Eds.), dHealth 2019 - From eHealth to dHelath: Proceedings of the 13th Health Informatics Meets Digital Health Conference (pp. 89-96). (Studies in health technology and informatics; Vol. 260). IOS Press. https://doi.org/10.3233/978-1-61499-971-3-89

Ensemble Based Approach for Time Series Classification in Metabolomics. / Netzer, Michael; Hanser, Friedrich; Breit, Marc; Weinberger, K.M.; Baumgartner, Christian; Baumgarten, Daniel.

dHealth 2019 - From eHealth to dHelath: Proceedings of the 13th Health Informatics Meets Digital Health Conference. ed. / Dieter Hayn; Alphons Eggerth; Günter Schreier. IOS Press, 2019. p. 89-96 (Studies in health technology and informatics; Vol. 260).

Research output: Chapter in Book/Report/Conference proceedingChapterResearchpeer-review

Netzer, M, Hanser, F, Breit, M, Weinberger, KM, Baumgartner, C & Baumgarten, D 2019, Ensemble Based Approach for Time Series Classification in Metabolomics. in D Hayn, A Eggerth & G Schreier (eds), dHealth 2019 - From eHealth to dHelath: Proceedings of the 13th Health Informatics Meets Digital Health Conference. Studies in health technology and informatics, vol. 260, IOS Press, pp. 89-96, dHealth 2019, Wien, Austria, 28/03/19. https://doi.org/10.3233/978-1-61499-971-3-89
Netzer M, Hanser F, Breit M, Weinberger KM, Baumgartner C, Baumgarten D. Ensemble Based Approach for Time Series Classification in Metabolomics. In Hayn D, Eggerth A, Schreier G, editors, dHealth 2019 - From eHealth to dHelath: Proceedings of the 13th Health Informatics Meets Digital Health Conference. IOS Press. 2019. p. 89-96. (Studies in health technology and informatics). https://doi.org/10.3233/978-1-61499-971-3-89
Netzer, Michael ; Hanser, Friedrich ; Breit, Marc ; Weinberger, K.M. ; Baumgartner, Christian ; Baumgarten, Daniel. / Ensemble Based Approach for Time Series Classification in Metabolomics. dHealth 2019 - From eHealth to dHelath: Proceedings of the 13th Health Informatics Meets Digital Health Conference. editor / Dieter Hayn ; Alphons Eggerth ; Günter Schreier. IOS Press, 2019. pp. 89-96 (Studies in health technology and informatics).
@inbook{979b6eb036bc41a4857be40cd1f9dd4b,
title = "Ensemble Based Approach for Time Series Classification in Metabolomics",
abstract = "BACKGROUND: Machine learning is one important application in the area of health informatics, however classification methods for longitudinal data are still rare. OBJECTIVES: The aim of this work is to analyze and classify differences in metabolite time series data between groups of individuals regarding their athletic activity. METHODS: We propose a new ensemble-based 2-tier approach to classify metabolite time series data. The first tier uses polynomial fitting to generate a class prediction for each metabolite. An induced classifier (k-nearest-neighbor or na{\"i}ve bayes) combines the results to produce a final prediction. Metabolite levels of 47 individuals undergoing a cycle ergometry test were measured using mass spectrometry. RESULTS: In accordance with our previous work the statistical results indicate strong changes over time. We found only small but systematic differences between the groups. However, our proposed stacking approach obtained a mean accuracy of 78{\%} using 10-fold cross-validation. CONCLUSION: Our proposed classification approach allows a considerable classification performance for time series data with small differences between the groups.",
keywords = "biomarkers, classification, kinetics, time series",
author = "Michael Netzer and Friedrich Hanser and Marc Breit and K.M. Weinberger and Christian Baumgartner and Daniel Baumgarten",
year = "2019",
doi = "10.3233/978-1-61499-971-3-89",
language = "English",
isbn = "978-1-61499-974-4",
series = "Studies in health technology and informatics",
publisher = "IOS Press",
pages = "89--96",
editor = "Dieter Hayn and Alphons Eggerth and G{\"u}nter Schreier",
booktitle = "dHealth 2019 - From eHealth to dHelath",
address = "Netherlands",

}

TY - CHAP

T1 - Ensemble Based Approach for Time Series Classification in Metabolomics

AU - Netzer, Michael

AU - Hanser, Friedrich

AU - Breit, Marc

AU - Weinberger, K.M.

AU - Baumgartner, Christian

AU - Baumgarten, Daniel

PY - 2019

Y1 - 2019

N2 - BACKGROUND: Machine learning is one important application in the area of health informatics, however classification methods for longitudinal data are still rare. OBJECTIVES: The aim of this work is to analyze and classify differences in metabolite time series data between groups of individuals regarding their athletic activity. METHODS: We propose a new ensemble-based 2-tier approach to classify metabolite time series data. The first tier uses polynomial fitting to generate a class prediction for each metabolite. An induced classifier (k-nearest-neighbor or naïve bayes) combines the results to produce a final prediction. Metabolite levels of 47 individuals undergoing a cycle ergometry test were measured using mass spectrometry. RESULTS: In accordance with our previous work the statistical results indicate strong changes over time. We found only small but systematic differences between the groups. However, our proposed stacking approach obtained a mean accuracy of 78% using 10-fold cross-validation. CONCLUSION: Our proposed classification approach allows a considerable classification performance for time series data with small differences between the groups.

AB - BACKGROUND: Machine learning is one important application in the area of health informatics, however classification methods for longitudinal data are still rare. OBJECTIVES: The aim of this work is to analyze and classify differences in metabolite time series data between groups of individuals regarding their athletic activity. METHODS: We propose a new ensemble-based 2-tier approach to classify metabolite time series data. The first tier uses polynomial fitting to generate a class prediction for each metabolite. An induced classifier (k-nearest-neighbor or naïve bayes) combines the results to produce a final prediction. Metabolite levels of 47 individuals undergoing a cycle ergometry test were measured using mass spectrometry. RESULTS: In accordance with our previous work the statistical results indicate strong changes over time. We found only small but systematic differences between the groups. However, our proposed stacking approach obtained a mean accuracy of 78% using 10-fold cross-validation. CONCLUSION: Our proposed classification approach allows a considerable classification performance for time series data with small differences between the groups.

KW - biomarkers

KW - classification

KW - kinetics

KW - time series

UR - https://graz.pure.elsevier.com/en/publications/ensemble-based-approach-for-time-series-classification-in-metabolomics(979b6eb0-36bc-41a4-857b-e40cd1f9dd4b).html

UR - http://www.scopus.com/inward/record.url?scp=85066474556&partnerID=8YFLogxK

U2 - 10.3233/978-1-61499-971-3-89

DO - 10.3233/978-1-61499-971-3-89

M3 - Chapter

SN - 978-1-61499-974-4

T3 - Studies in health technology and informatics

SP - 89

EP - 96

BT - dHealth 2019 - From eHealth to dHelath

A2 - Hayn, Dieter

A2 - Eggerth, Alphons

A2 - Schreier, Günter

PB - IOS Press

ER -