A probabilistic framework for behavioral identification from animal-borne accelerometers

Dentinger, Jane E.; Borger, Luca; Holton, Mark; Jafari-Marandi, Ruholla; Norman, Durham A.; Smith, Brian K.; Oppenheimer, Seth F.; Strickland, Bronson K.; Wilson, Rory; Street, Garrett M.

doi:10.1016/j.ecolmodel.2021.109818

Journal article 824 views 136 downloads

A probabilistic framework for behavioral identification from animal-borne accelerometers

Jane E. Dentinger, Luca Borger

, Mark Holton

, Ruholla Jafari-Marandi, Durham A. Norman, Brian K. Smith, Seth F. Oppenheimer, Bronson K. Strickland, Rory Wilson

, Garrett M. Street

Ecological Modelling, Volume: 464, Start page: 109818

Swansea University Authors: Luca Borger , Mark Holton , Rory Wilson

PDF | Accepted Manuscript

©2021 All rights reserved. All article content, except where otherwise noted, is licensed under a Creative Commons Attribution Non-Commercial No Derivatives License (CC-BY-NC-ND)
Download (971.88KB)

Check full text

DOI (Published version): 10.1016/j.ecolmodel.2021.109818

Abstract

Many studies of animal distributions use habitat and climactic variables to explain patterns of observed space use. However, without behavioral information, we can only speculate as to why and how these characteristics are important to species persistence. Animal-borne accelerometer and magnetometer...

Full description

Published in:	Ecological Modelling
ISSN:	0304-3800
Published:	Elsevier BV 2022
Online Access:	Check full text
URI:	https://cronfa.swan.ac.uk/Record/cronfa58741
Tags:	Add Tag No Tags, Be the first to tag this record!

first_indexed	2021-11-22T13:17:13Z
last_indexed	2021-12-07T04:15:39Z
id	cronfa58741
recordtype	SURis
fullrecord	<?xml version="1.0"?><rfc1807><datestamp>2021-12-06T16:24:24.6811929</datestamp><bib-version>v2</bib-version><id>58741</id><entry>2021-11-22</entry><title>A probabilistic framework for behavioral identification from animal-borne accelerometers</title><swanseaauthors><author><sid>8416d0ffc3cccdad6e6d67a455e7c4a2</sid><ORCID>0000-0001-8763-5997</ORCID><firstname>Luca</firstname><surname>Borger</surname><name>Luca Borger</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>0e1d89d0cc934a740dcd0a873aed178e</sid><ORCID>0000-0001-8834-3283</ORCID><firstname>Mark</firstname><surname>Holton</surname><name>Mark Holton</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>017bc6dd155098860945dc6249c4e9bc</sid><ORCID>0000-0003-3177-0177</ORCID><firstname>Rory</firstname><surname>Wilson</surname><name>Rory Wilson</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2021-11-22</date><deptcode>SBI</deptcode><abstract>Many studies of animal distributions use habitat and climactic variables to explain patterns of observed space use. However, without behavioral information, we can only speculate as to why and how these characteristics are important to species persistence. Animal-borne accelerometer and magnetometer data loggers can be used to detect behaviors and when coupled with telemetry improve our understanding of animal space use and habitat requirements. However, these loggers collect tremendous quantities of data requiring automated machine learning techniques to identify patterns in the data. Supervised machine learning requires a set of training signals with known behaviors to train the model to identify the unique signal characteristics associated with each behavior. In contrast, unsupervised approaches aggregate unlabeled signals into groups based purely on signal similarity but, without additional information, do not identify specific behaviors. In this paper, we propose a probabilistic framework for interpreting uncertainty in machine learning techniques—the probability profile—and demonstrate how to post hoc identify behaviors within signal groups. We assess model performance using a matrix-based measure of dissimilarity. We used a Random Forest (RF) and a clustered self-organizing map (CSOM) for comparison and demonstrate the use of a behavioral profile for each using a data set of high-frequency accelerometer and magnetometer data collected from 7 captive wild pigs (Sus scrofa) moving in a 1 ha outdoor enclosure. We found that the RF had more discrimination than the CSOM which had fewer clusters associated with high probabilities of a single behavior (>50%). The leave-p-out cross validation statistic of the probability matrix ( L 1 ¯ ) indicated that there was an average maximum dissimilarity of 20% and 65% between the training and test data sets for the RF and CSOM methods, respectively. Using a probability profile to describe groups predicted from machine learning allows the variation and error inherent in behavioral prediction to be incorporated directly into the model to better reflect the nuances of behavior derived from accelerometer and/or magnetometer signals. We discuss the data requirements of this framework, demonstrate its application to field data, highlight critical assumptions and caveats, and examine how it may be used to generate new ecological inference.</abstract><type>Journal Article</type><journal>Ecological Modelling</journal><volume>464</volume><journalNumber/><paginationStart>109818</paginationStart><paginationEnd/><publisher>Elsevier BV</publisher><placeOfPublication/><isbnPrint/><isbnElectronic/><issnPrint>0304-3800</issnPrint><issnElectronic/><keywords>Accelerometers; Behavior; Machine learning; k-means clustering; SOM; Random forest</keywords><publishedDay>1</publishedDay><publishedMonth>2</publishedMonth><publishedYear>2022</publishedYear><publishedDate>2022-02-01</publishedDate><doi>10.1016/j.ecolmodel.2021.109818</doi><url/><notes/><college>COLLEGE NANME</college><department>Biosciences</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>SBI</DepartmentCode><institution>Swansea University</institution><apcterm/><funders>Mississippi Agriculture and Forestry Research Station Strategic Research Initiative Noble Research Institute</funders><lastEdited>2021-12-06T16:24:24.6811929</lastEdited><Created>2021-11-22T13:13:42.3130708</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Biosciences, Geography and Physics - Biosciences</level></path><authors><author><firstname>Jane E.</firstname><surname>Dentinger</surname><order>1</order></author><author><firstname>Luca</firstname><surname>Borger</surname><orcid>0000-0001-8763-5997</orcid><order>2</order></author><author><firstname>Mark</firstname><surname>Holton</surname><orcid>0000-0001-8834-3283</orcid><order>3</order></author><author><firstname>Ruholla</firstname><surname>Jafari-Marandi</surname><order>4</order></author><author><firstname>Durham A.</firstname><surname>Norman</surname><order>5</order></author><author><firstname>Brian K.</firstname><surname>Smith</surname><order>6</order></author><author><firstname>Seth F.</firstname><surname>Oppenheimer</surname><order>7</order></author><author><firstname>Bronson K.</firstname><surname>Strickland</surname><order>8</order></author><author><firstname>Rory</firstname><surname>Wilson</surname><orcid>0000-0003-3177-0177</orcid><order>9</order></author><author><firstname>Garrett M.</firstname><surname>Street</surname><order>10</order></author></authors><documents><document><filename>58741__21770__b0ec8512a2d348abaeb678e65713e9ae.pdf</filename><originalFilename>58741.pdf</originalFilename><uploaded>2021-12-02T10:29:09.5733483</uploaded><type>Output</type><contentLength>995208</contentLength><contentType>application/pdf</contentType><version>Accepted Manuscript</version><cronfaStatus>true</cronfaStatus><embargoDate>2022-11-14T00:00:00.0000000</embargoDate><documentNotes>©2021 All rights reserved. All article content, except where otherwise noted, is licensed under a Creative Commons Attribution Non-Commercial No Derivatives License (CC-BY-NC-ND)</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>https://www.creativecommons.org/licenses/by-nc-nd/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling	2021-12-06T16:24:24.6811929 v2 58741 2021-11-22 A probabilistic framework for behavioral identification from animal-borne accelerometers 8416d0ffc3cccdad6e6d67a455e7c4a2 0000-0001-8763-5997 Luca Borger Luca Borger true false 0e1d89d0cc934a740dcd0a873aed178e 0000-0001-8834-3283 Mark Holton Mark Holton true false 017bc6dd155098860945dc6249c4e9bc 0000-0003-3177-0177 Rory Wilson Rory Wilson true false 2021-11-22 SBI Many studies of animal distributions use habitat and climactic variables to explain patterns of observed space use. However, without behavioral information, we can only speculate as to why and how these characteristics are important to species persistence. Animal-borne accelerometer and magnetometer data loggers can be used to detect behaviors and when coupled with telemetry improve our understanding of animal space use and habitat requirements. However, these loggers collect tremendous quantities of data requiring automated machine learning techniques to identify patterns in the data. Supervised machine learning requires a set of training signals with known behaviors to train the model to identify the unique signal characteristics associated with each behavior. In contrast, unsupervised approaches aggregate unlabeled signals into groups based purely on signal similarity but, without additional information, do not identify specific behaviors. In this paper, we propose a probabilistic framework for interpreting uncertainty in machine learning techniques—the probability profile—and demonstrate how to post hoc identify behaviors within signal groups. We assess model performance using a matrix-based measure of dissimilarity. We used a Random Forest (RF) and a clustered self-organizing map (CSOM) for comparison and demonstrate the use of a behavioral profile for each using a data set of high-frequency accelerometer and magnetometer data collected from 7 captive wild pigs (Sus scrofa) moving in a 1 ha outdoor enclosure. We found that the RF had more discrimination than the CSOM which had fewer clusters associated with high probabilities of a single behavior (>50%). The leave-p-out cross validation statistic of the probability matrix ( L 1 ¯ ) indicated that there was an average maximum dissimilarity of 20% and 65% between the training and test data sets for the RF and CSOM methods, respectively. Using a probability profile to describe groups predicted from machine learning allows the variation and error inherent in behavioral prediction to be incorporated directly into the model to better reflect the nuances of behavior derived from accelerometer and/or magnetometer signals. We discuss the data requirements of this framework, demonstrate its application to field data, highlight critical assumptions and caveats, and examine how it may be used to generate new ecological inference. Journal Article Ecological Modelling 464 109818 Elsevier BV 0304-3800 Accelerometers; Behavior; Machine learning; k-means clustering; SOM; Random forest 1 2 2022 2022-02-01 10.1016/j.ecolmodel.2021.109818 COLLEGE NANME Biosciences COLLEGE CODE SBI Swansea University Mississippi Agriculture and Forestry Research Station Strategic Research Initiative Noble Research Institute 2021-12-06T16:24:24.6811929 2021-11-22T13:13:42.3130708 Faculty of Science and Engineering School of Biosciences, Geography and Physics - Biosciences Jane E. Dentinger 1 Luca Borger 0000-0001-8763-5997 2 Mark Holton 0000-0001-8834-3283 3 Ruholla Jafari-Marandi 4 Durham A. Norman 5 Brian K. Smith 6 Seth F. Oppenheimer 7 Bronson K. Strickland 8 Rory Wilson 0000-0003-3177-0177 9 Garrett M. Street 10 58741__21770__b0ec8512a2d348abaeb678e65713e9ae.pdf 58741.pdf 2021-12-02T10:29:09.5733483 Output 995208 application/pdf Accepted Manuscript true 2022-11-14T00:00:00.0000000 ©2021 All rights reserved. All article content, except where otherwise noted, is licensed under a Creative Commons Attribution Non-Commercial No Derivatives License (CC-BY-NC-ND) true eng https://www.creativecommons.org/licenses/by-nc-nd/4.0/
title	A probabilistic framework for behavioral identification from animal-borne accelerometers
spellingShingle	A probabilistic framework for behavioral identification from animal-borne accelerometers Luca Borger Mark Holton Rory Wilson
title_short	A probabilistic framework for behavioral identification from animal-borne accelerometers
title_full	A probabilistic framework for behavioral identification from animal-borne accelerometers
title_fullStr	A probabilistic framework for behavioral identification from animal-borne accelerometers
title_full_unstemmed	A probabilistic framework for behavioral identification from animal-borne accelerometers
title_sort	A probabilistic framework for behavioral identification from animal-borne accelerometers
author_id_str_mv	8416d0ffc3cccdad6e6d67a455e7c4a2 0e1d89d0cc934a740dcd0a873aed178e 017bc6dd155098860945dc6249c4e9bc
author_id_fullname_str_mv	8416d0ffc3cccdad6e6d67a455e7c4a2_*_Luca Borger 0e1d89d0cc934a740dcd0a873aed178e__Mark Holton 017bc6dd155098860945dc6249c4e9bc_**_Rory Wilson
author	Luca Borger Mark Holton Rory Wilson
author2	Jane E. Dentinger Luca Borger Mark Holton Ruholla Jafari-Marandi Durham A. Norman Brian K. Smith Seth F. Oppenheimer Bronson K. Strickland Rory Wilson Garrett M. Street
format	Journal article
container_title	Ecological Modelling
container_volume	464
container_start_page	109818
publishDate	2022
institution	Swansea University
issn	0304-3800
doi_str_mv	10.1016/j.ecolmodel.2021.109818
publisher	Elsevier BV
college_str	Faculty of Science and Engineering
hierarchytype
hierarchy_top_id	facultyofscienceandengineering
hierarchy_top_title	Faculty of Science and Engineering
hierarchy_parent_id	facultyofscienceandengineering
hierarchy_parent_title	Faculty of Science and Engineering
department_str	School of Biosciences, Geography and Physics - Biosciences{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Biosciences, Geography and Physics - Biosciences
document_store_str	1
active_str	0
description	Many studies of animal distributions use habitat and climactic variables to explain patterns of observed space use. However, without behavioral information, we can only speculate as to why and how these characteristics are important to species persistence. Animal-borne accelerometer and magnetometer data loggers can be used to detect behaviors and when coupled with telemetry improve our understanding of animal space use and habitat requirements. However, these loggers collect tremendous quantities of data requiring automated machine learning techniques to identify patterns in the data. Supervised machine learning requires a set of training signals with known behaviors to train the model to identify the unique signal characteristics associated with each behavior. In contrast, unsupervised approaches aggregate unlabeled signals into groups based purely on signal similarity but, without additional information, do not identify specific behaviors. In this paper, we propose a probabilistic framework for interpreting uncertainty in machine learning techniques—the probability profile—and demonstrate how to post hoc identify behaviors within signal groups. We assess model performance using a matrix-based measure of dissimilarity. We used a Random Forest (RF) and a clustered self-organizing map (CSOM) for comparison and demonstrate the use of a behavioral profile for each using a data set of high-frequency accelerometer and magnetometer data collected from 7 captive wild pigs (Sus scrofa) moving in a 1 ha outdoor enclosure. We found that the RF had more discrimination than the CSOM which had fewer clusters associated with high probabilities of a single behavior (>50%). The leave-p-out cross validation statistic of the probability matrix ( L 1 ¯ ) indicated that there was an average maximum dissimilarity of 20% and 65% between the training and test data sets for the RF and CSOM methods, respectively. Using a probability profile to describe groups predicted from machine learning allows the variation and error inherent in behavioral prediction to be incorporated directly into the model to better reflect the nuances of behavior derived from accelerometer and/or magnetometer signals. We discuss the data requirements of this framework, demonstrate its application to field data, highlight critical assumptions and caveats, and examine how it may be used to generate new ecological inference.
published_date	2022-02-01T04:15:31Z
_version_	1763754042774781952
score	11.035634

A probabilistic framework for behavioral identification from animal-borne accelerometers

Similar Items