The use of an unsupervised learning approach for characterizing latent behaviors in accelerometer data

Marianna Chimienti, Thomas Cornulier, Ellie Owen, Mark Bolton, Ian M. Davies, Justin M. J. Travis, Beth E. Scott

Research output: Contribution to journalArticle

22 Citations (Scopus)
4 Downloads (Pure)

Abstract

The recent increase in data accuracy from high resolution accelerometers offers substantial potential for improved understanding and prediction of animal movements. However, current approaches used for analysing these multivariable
datasets typically require existing knowledge of the behaviors of the animals to inform the behavioral classification process. These methods are thus not wellsuited for the many cases where limited knowledge of the different behaviors performed exist. Here, we introduce the use of an unsupervised learning algorithm. To illustrate the method’s capability we analyse data collected using a combination of GPS and Accelerometers on two seabird species: razorbills (Alca torda) and common guillemots (Uria aalge). We applied the unsupervised learning algorithm Expectation Maximization to characterize latent behavioral
states both above and below water at both individual and group level. The application of this flexible approach yielded significant new insights into the foraging strategies of the two study species, both above and below the surface of the water. In addition to general behavioral modes such as flying, floating, as well as descending and ascending phases within the water column, this
approach allowed an exploration of previously unstudied and important behaviors such as searching and prey chasing/capture events. We propose that this unsupervised learning approach provides an ideal tool for the systematic analysis of such complex multivariable movement data that are increasingly being obtained with accelerometer tags across species. In particular, we recommend its application in cases where we have limited current knowledge of the behaviors performed and existing supervised learning approaches may have limited utility.
Original languageEnglish
Pages (from-to)727–741
Number of pages15
JournalEcology and Evolution
Volume6
Issue number3
Early online date11 Jan 2016
DOIs
Publication statusPublished - Feb 2016

Fingerprint

accelerometer
learning
Alcidae
prey capture
animal
seabird
seabirds
animal behavior
surface water
GPS
flight
water
water column
foraging
taxonomy
prediction
methodology
animals
method

Keywords

  • Accelerometer data
  • animal movements
  • behavioral classification
  • unsupervised learning

Cite this

The use of an unsupervised learning approach for characterizing latent behaviors in accelerometer data. / Chimienti, Marianna; Cornulier, Thomas; Owen, Ellie; Bolton, Mark; Davies, Ian M.; Travis, Justin M. J.; Scott, Beth E.

In: Ecology and Evolution, Vol. 6, No. 3, 02.2016, p. 727–741.

Research output: Contribution to journalArticle

@article{6ea221f6cc7c417393ba7d556e768ade,
title = "The use of an unsupervised learning approach for characterizing latent behaviors in accelerometer data",
abstract = "The recent increase in data accuracy from high resolution accelerometers offers substantial potential for improved understanding and prediction of animal movements. However, current approaches used for analysing these multivariabledatasets typically require existing knowledge of the behaviors of the animals to inform the behavioral classification process. These methods are thus not wellsuited for the many cases where limited knowledge of the different behaviors performed exist. Here, we introduce the use of an unsupervised learning algorithm. To illustrate the method’s capability we analyse data collected using a combination of GPS and Accelerometers on two seabird species: razorbills (Alca torda) and common guillemots (Uria aalge). We applied the unsupervised learning algorithm Expectation Maximization to characterize latent behavioralstates both above and below water at both individual and group level. The application of this flexible approach yielded significant new insights into the foraging strategies of the two study species, both above and below the surface of the water. In addition to general behavioral modes such as flying, floating, as well as descending and ascending phases within the water column, thisapproach allowed an exploration of previously unstudied and important behaviors such as searching and prey chasing/capture events. We propose that this unsupervised learning approach provides an ideal tool for the systematic analysis of such complex multivariable movement data that are increasingly being obtained with accelerometer tags across species. In particular, we recommend its application in cases where we have limited current knowledge of the behaviors performed and existing supervised learning approaches may have limited utility.",
keywords = "Accelerometer data, animal movements, behavioral classification, unsupervised learning",
author = "Marianna Chimienti and Thomas Cornulier and Ellie Owen and Mark Bolton and Davies, {Ian M.} and Travis, {Justin M. J.} and Scott, {Beth E.}",
note = "Acknowledgments This project and the tags deployed on both seabird's species were fund by NERC (grant number NE/K007440/1), Marine Scotland Science and Seabird Tracking and Research (STAR) Project led by the Royal Society for the Protection of Birds (RSPB). We would like to thank Rob Hughes, Tessa Cole and Ruth Brown for helping in the data collection, the Bird Observatory of Fair Isle for supporting the fieldwork and the Marine Collaboration Research Forum (MarCRF).",
year = "2016",
month = "2",
doi = "10.1002/ece3.1914",
language = "English",
volume = "6",
pages = "727–741",
journal = "Ecology and Evolution",
issn = "2045-7758",
publisher = "Wiley",
number = "3",

}

TY - JOUR

T1 - The use of an unsupervised learning approach for characterizing latent behaviors in accelerometer data

AU - Chimienti, Marianna

AU - Cornulier, Thomas

AU - Owen, Ellie

AU - Bolton, Mark

AU - Davies, Ian M.

AU - Travis, Justin M. J.

AU - Scott, Beth E.

N1 - Acknowledgments This project and the tags deployed on both seabird's species were fund by NERC (grant number NE/K007440/1), Marine Scotland Science and Seabird Tracking and Research (STAR) Project led by the Royal Society for the Protection of Birds (RSPB). We would like to thank Rob Hughes, Tessa Cole and Ruth Brown for helping in the data collection, the Bird Observatory of Fair Isle for supporting the fieldwork and the Marine Collaboration Research Forum (MarCRF).

PY - 2016/2

Y1 - 2016/2

N2 - The recent increase in data accuracy from high resolution accelerometers offers substantial potential for improved understanding and prediction of animal movements. However, current approaches used for analysing these multivariabledatasets typically require existing knowledge of the behaviors of the animals to inform the behavioral classification process. These methods are thus not wellsuited for the many cases where limited knowledge of the different behaviors performed exist. Here, we introduce the use of an unsupervised learning algorithm. To illustrate the method’s capability we analyse data collected using a combination of GPS and Accelerometers on two seabird species: razorbills (Alca torda) and common guillemots (Uria aalge). We applied the unsupervised learning algorithm Expectation Maximization to characterize latent behavioralstates both above and below water at both individual and group level. The application of this flexible approach yielded significant new insights into the foraging strategies of the two study species, both above and below the surface of the water. In addition to general behavioral modes such as flying, floating, as well as descending and ascending phases within the water column, thisapproach allowed an exploration of previously unstudied and important behaviors such as searching and prey chasing/capture events. We propose that this unsupervised learning approach provides an ideal tool for the systematic analysis of such complex multivariable movement data that are increasingly being obtained with accelerometer tags across species. In particular, we recommend its application in cases where we have limited current knowledge of the behaviors performed and existing supervised learning approaches may have limited utility.

AB - The recent increase in data accuracy from high resolution accelerometers offers substantial potential for improved understanding and prediction of animal movements. However, current approaches used for analysing these multivariabledatasets typically require existing knowledge of the behaviors of the animals to inform the behavioral classification process. These methods are thus not wellsuited for the many cases where limited knowledge of the different behaviors performed exist. Here, we introduce the use of an unsupervised learning algorithm. To illustrate the method’s capability we analyse data collected using a combination of GPS and Accelerometers on two seabird species: razorbills (Alca torda) and common guillemots (Uria aalge). We applied the unsupervised learning algorithm Expectation Maximization to characterize latent behavioralstates both above and below water at both individual and group level. The application of this flexible approach yielded significant new insights into the foraging strategies of the two study species, both above and below the surface of the water. In addition to general behavioral modes such as flying, floating, as well as descending and ascending phases within the water column, thisapproach allowed an exploration of previously unstudied and important behaviors such as searching and prey chasing/capture events. We propose that this unsupervised learning approach provides an ideal tool for the systematic analysis of such complex multivariable movement data that are increasingly being obtained with accelerometer tags across species. In particular, we recommend its application in cases where we have limited current knowledge of the behaviors performed and existing supervised learning approaches may have limited utility.

KW - Accelerometer data

KW - animal movements

KW - behavioral classification

KW - unsupervised learning

U2 - 10.1002/ece3.1914

DO - 10.1002/ece3.1914

M3 - Article

VL - 6

SP - 727

EP - 741

JO - Ecology and Evolution

JF - Ecology and Evolution

SN - 2045-7758

IS - 3

ER -