Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach
Autor(a) principal: | |
---|---|
Data de Publicação: | 2018 |
Outros Autores: | , , , , , |
Tipo de documento: | Artigo |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UNESP |
Texto Completo: | http://dx.doi.org/10.1007/s11265-018-1376-5 http://hdl.handle.net/11449/179835 |
Resumo: | Multi-target regression (MTR) regards predictive problems with multiple numerical targets. To solve this, machine learning techniques can model solutions treating each target as a separated problem based only on the input features. Nonetheless, modelling inter-target correlation can improve predictive performance. When performing MTR tasks using the statistical dependencies of targets, several approaches put aside the evaluation of each pair-wise correlation between those targets, which may differ for each problem. Besides that, one of the main drawbacks of the current leading MTR method is its high memory cost. In this paper, we propose a novel MTR method called Multi-output Tree Chaining (MOTC) to overcome the mentioned disadvantages. Our method provides an interpretative internal tree-based structure which represents the relationships between targets denominated Chaining Trees (CT). Different from the current techniques, we compute the outputs dependencies, one-by-one, based on the Random Forest importance metric. Furthermore, we proposed a memory friendly approach which reduces the number of required regression models when compared to a leading method, reducing computational cost. We compared the proposed algorithm against three MTR methods (Single-target - ST; Multi-Target Regressor Stacking - MTRS; and Ensemble of Regressor Chains - ERC) on 18 benchmark datasets with two base regression algorithms (Random Forest and Support Vector Regression). The obtained results show that our method is superior to the ST approach regarding predictive performance, whereas, having no significant difference from ERC and MTRS. Moreover, the interpretative tree-based structures built by MOTC pose as great insight on the relationships among targets. Lastly, the proposed solution used significantly less memory than ERC being very similar in predictive performance. |
id |
UNSP_5bb23e1ef000c15e32f3441bc33d5319 |
---|---|
oai_identifier_str |
oai:repositorio.unesp.br:11449/179835 |
network_acronym_str |
UNSP |
network_name_str |
Repositório Institucional da UNESP |
repository_id_str |
2946 |
spelling |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target ApproachInterpretative tree structureMachine learningMemory-friendly algorithmMulti-outputMulti-target regressionMulti-target regression (MTR) regards predictive problems with multiple numerical targets. To solve this, machine learning techniques can model solutions treating each target as a separated problem based only on the input features. Nonetheless, modelling inter-target correlation can improve predictive performance. When performing MTR tasks using the statistical dependencies of targets, several approaches put aside the evaluation of each pair-wise correlation between those targets, which may differ for each problem. Besides that, one of the main drawbacks of the current leading MTR method is its high memory cost. In this paper, we propose a novel MTR method called Multi-output Tree Chaining (MOTC) to overcome the mentioned disadvantages. Our method provides an interpretative internal tree-based structure which represents the relationships between targets denominated Chaining Trees (CT). Different from the current techniques, we compute the outputs dependencies, one-by-one, based on the Random Forest importance metric. Furthermore, we proposed a memory friendly approach which reduces the number of required regression models when compared to a leading method, reducing computational cost. We compared the proposed algorithm against three MTR methods (Single-target - ST; Multi-Target Regressor Stacking - MTRS; and Ensemble of Regressor Chains - ERC) on 18 benchmark datasets with two base regression algorithms (Random Forest and Support Vector Regression). The obtained results show that our method is superior to the ST approach regarding predictive performance, whereas, having no significant difference from ERC and MTRS. Moreover, the interpretative tree-based structures built by MOTC pose as great insight on the relationships among targets. Lastly, the proposed solution used significantly less memory than ERC being very similar in predictive performance.Computer Science Department State University of Londrina. Rodovia Celso Garcia Cid Km 380 s/n - Campus UniversitárioElectrical Engineering Department State University of Londrina. Rodovia Celso Garcia Cid Km 380 s/n - Campus UniversitárioDepartment of Computer Science Federal University of São Carlos Rodovia Washington Luís, km 235Instituto de Biociências Letras e Ciências Exatas Unesp - Univ Estadual Paulista (São Paulo State University), Rua Cristóvão Colombo 2265, Jd NazarethInstituto de Biociências Letras e Ciências Exatas Unesp - Univ Estadual Paulista (São Paulo State University), Rua Cristóvão Colombo 2265, Jd NazarethUniversidade Estadual de Londrina (UEL)Universidade Federal de São Carlos (UFSCar)Universidade Estadual Paulista (Unesp)Mastelini, Saulo Martielloda Costa, Victor Guilherme TurrisiSantana, Everton JoseNakano, Felipe KenjiGuido, Rodrigo Capobianco [UNESP]Cerri, RicardoBarbon, Sylvio2018-12-11T17:36:57Z2018-12-11T17:36:57Z2018-05-05info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/article1-25application/pdfhttp://dx.doi.org/10.1007/s11265-018-1376-5Journal of Signal Processing Systems, p. 1-25.1939-81151939-8018http://hdl.handle.net/11449/17983510.1007/s11265-018-1376-52-s2.0-850464962142-s2.0-85046496214.pdf65420862268080670000-0002-0924-8024Scopusreponame:Repositório Institucional da UNESPinstname:Universidade Estadual Paulista (UNESP)instacron:UNESPengJournal of Signal Processing Systems0,216info:eu-repo/semantics/openAccess2023-10-29T06:04:45Zoai:repositorio.unesp.br:11449/179835Repositório InstitucionalPUBhttp://repositorio.unesp.br/oai/requestopendoar:29462024-08-05T16:19:21.740943Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP)false |
dc.title.none.fl_str_mv |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
title |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
spellingShingle |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach Mastelini, Saulo Martiello Interpretative tree structure Machine learning Memory-friendly algorithm Multi-output Multi-target regression |
title_short |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
title_full |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
title_fullStr |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
title_full_unstemmed |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
title_sort |
Multi-Output Tree Chaining: An Interpretative Modelling and Lightweight Multi-Target Approach |
author |
Mastelini, Saulo Martiello |
author_facet |
Mastelini, Saulo Martiello da Costa, Victor Guilherme Turrisi Santana, Everton Jose Nakano, Felipe Kenji Guido, Rodrigo Capobianco [UNESP] Cerri, Ricardo Barbon, Sylvio |
author_role |
author |
author2 |
da Costa, Victor Guilherme Turrisi Santana, Everton Jose Nakano, Felipe Kenji Guido, Rodrigo Capobianco [UNESP] Cerri, Ricardo Barbon, Sylvio |
author2_role |
author author author author author author |
dc.contributor.none.fl_str_mv |
Universidade Estadual de Londrina (UEL) Universidade Federal de São Carlos (UFSCar) Universidade Estadual Paulista (Unesp) |
dc.contributor.author.fl_str_mv |
Mastelini, Saulo Martiello da Costa, Victor Guilherme Turrisi Santana, Everton Jose Nakano, Felipe Kenji Guido, Rodrigo Capobianco [UNESP] Cerri, Ricardo Barbon, Sylvio |
dc.subject.por.fl_str_mv |
Interpretative tree structure Machine learning Memory-friendly algorithm Multi-output Multi-target regression |
topic |
Interpretative tree structure Machine learning Memory-friendly algorithm Multi-output Multi-target regression |
description |
Multi-target regression (MTR) regards predictive problems with multiple numerical targets. To solve this, machine learning techniques can model solutions treating each target as a separated problem based only on the input features. Nonetheless, modelling inter-target correlation can improve predictive performance. When performing MTR tasks using the statistical dependencies of targets, several approaches put aside the evaluation of each pair-wise correlation between those targets, which may differ for each problem. Besides that, one of the main drawbacks of the current leading MTR method is its high memory cost. In this paper, we propose a novel MTR method called Multi-output Tree Chaining (MOTC) to overcome the mentioned disadvantages. Our method provides an interpretative internal tree-based structure which represents the relationships between targets denominated Chaining Trees (CT). Different from the current techniques, we compute the outputs dependencies, one-by-one, based on the Random Forest importance metric. Furthermore, we proposed a memory friendly approach which reduces the number of required regression models when compared to a leading method, reducing computational cost. We compared the proposed algorithm against three MTR methods (Single-target - ST; Multi-Target Regressor Stacking - MTRS; and Ensemble of Regressor Chains - ERC) on 18 benchmark datasets with two base regression algorithms (Random Forest and Support Vector Regression). The obtained results show that our method is superior to the ST approach regarding predictive performance, whereas, having no significant difference from ERC and MTRS. Moreover, the interpretative tree-based structures built by MOTC pose as great insight on the relationships among targets. Lastly, the proposed solution used significantly less memory than ERC being very similar in predictive performance. |
publishDate |
2018 |
dc.date.none.fl_str_mv |
2018-12-11T17:36:57Z 2018-12-11T17:36:57Z 2018-05-05 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/article |
format |
article |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://dx.doi.org/10.1007/s11265-018-1376-5 Journal of Signal Processing Systems, p. 1-25. 1939-8115 1939-8018 http://hdl.handle.net/11449/179835 10.1007/s11265-018-1376-5 2-s2.0-85046496214 2-s2.0-85046496214.pdf 6542086226808067 0000-0002-0924-8024 |
url |
http://dx.doi.org/10.1007/s11265-018-1376-5 http://hdl.handle.net/11449/179835 |
identifier_str_mv |
Journal of Signal Processing Systems, p. 1-25. 1939-8115 1939-8018 10.1007/s11265-018-1376-5 2-s2.0-85046496214 2-s2.0-85046496214.pdf 6542086226808067 0000-0002-0924-8024 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.relation.none.fl_str_mv |
Journal of Signal Processing Systems 0,216 |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
1-25 application/pdf |
dc.source.none.fl_str_mv |
Scopus reponame:Repositório Institucional da UNESP instname:Universidade Estadual Paulista (UNESP) instacron:UNESP |
instname_str |
Universidade Estadual Paulista (UNESP) |
instacron_str |
UNESP |
institution |
UNESP |
reponame_str |
Repositório Institucional da UNESP |
collection |
Repositório Institucional da UNESP |
repository.name.fl_str_mv |
Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP) |
repository.mail.fl_str_mv |
|
_version_ |
1808128632657281024 |