Explaining dimensionality reduction results using Shapley values
Autor(a) principal: | |
---|---|
Data de Publicação: | 2021 |
Outros Autores: | |
Tipo de documento: | Artigo |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UNESP |
Texto Completo: | http://dx.doi.org/10.1016/j.eswa.2021.115020 http://hdl.handle.net/11449/206269 |
Resumo: | Dimensionality reduction (DR) techniques have been consistently supporting high-dimensional data analysis in various applications. Besides the patterns uncovered by these techniques, the interpretation of DR results based on each feature's contribution to the low-dimensional representation supports new finds through exploratory analysis. Current literature approaches designed to interpret DR techniques do not explain the features’ contributions well since they focus only on the low-dimensional representation or do not consider the relationship among features. This paper presents ClusterShapley to address these problems, using Shapley values to generate explanations of dimensionality reduction techniques and interpret these algorithms using a cluster-oriented analysis. ClusterShapley explains the formation of clusters and the meaning of their relationship, which is useful for exploratory data analysis in various domains. We propose novel visualization techniques to guide the interpretation of features’ contributions on clustering formation and validate our methodology through case studies of publicly available datasets. The results demonstrate our approach's interpretability and analysis power to generate insights about pathologies and patients in different conditions using DR results. |
id |
UNSP_366fb4f7ecbd6b3909f628895d75d036 |
---|---|
oai_identifier_str |
oai:repositorio.unesp.br:11449/206269 |
network_acronym_str |
UNSP |
network_name_str |
Repositório Institucional da UNESP |
repository_id_str |
2946 |
spelling |
Explaining dimensionality reduction results using Shapley valuesDimensionality reductionExplainabilityShapley valuesVisualizationDimensionality reduction (DR) techniques have been consistently supporting high-dimensional data analysis in various applications. Besides the patterns uncovered by these techniques, the interpretation of DR results based on each feature's contribution to the low-dimensional representation supports new finds through exploratory analysis. Current literature approaches designed to interpret DR techniques do not explain the features’ contributions well since they focus only on the low-dimensional representation or do not consider the relationship among features. This paper presents ClusterShapley to address these problems, using Shapley values to generate explanations of dimensionality reduction techniques and interpret these algorithms using a cluster-oriented analysis. ClusterShapley explains the formation of clusters and the meaning of their relationship, which is useful for exploratory data analysis in various domains. We propose novel visualization techniques to guide the interpretation of features’ contributions on clustering formation and validate our methodology through case studies of publicly available datasets. The results demonstrate our approach's interpretability and analysis power to generate insights about pathologies and patients in different conditions using DR results.Faculty of Sciences and Technology São Paulo State University (UNESP)Faculty of Sciences and Technology São Paulo State University (UNESP)Universidade Estadual Paulista (Unesp)Marcílio-Jr, Wilson E. [UNESP]Eler, Danilo M. [UNESP]2021-06-25T10:29:21Z2021-06-25T10:29:21Z2021-09-15info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/articlehttp://dx.doi.org/10.1016/j.eswa.2021.115020Expert Systems with Applications, v. 178.0957-4174http://hdl.handle.net/11449/20626910.1016/j.eswa.2021.1150202-s2.0-85105034882Scopusreponame:Repositório Institucional da UNESPinstname:Universidade Estadual Paulista (UNESP)instacron:UNESPengExpert Systems with Applicationsinfo:eu-repo/semantics/openAccess2021-10-23T02:54:11Zoai:repositorio.unesp.br:11449/206269Repositório InstitucionalPUBhttp://repositorio.unesp.br/oai/requestopendoar:29462021-10-23T02:54:11Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP)false |
dc.title.none.fl_str_mv |
Explaining dimensionality reduction results using Shapley values |
title |
Explaining dimensionality reduction results using Shapley values |
spellingShingle |
Explaining dimensionality reduction results using Shapley values Marcílio-Jr, Wilson E. [UNESP] Dimensionality reduction Explainability Shapley values Visualization |
title_short |
Explaining dimensionality reduction results using Shapley values |
title_full |
Explaining dimensionality reduction results using Shapley values |
title_fullStr |
Explaining dimensionality reduction results using Shapley values |
title_full_unstemmed |
Explaining dimensionality reduction results using Shapley values |
title_sort |
Explaining dimensionality reduction results using Shapley values |
author |
Marcílio-Jr, Wilson E. [UNESP] |
author_facet |
Marcílio-Jr, Wilson E. [UNESP] Eler, Danilo M. [UNESP] |
author_role |
author |
author2 |
Eler, Danilo M. [UNESP] |
author2_role |
author |
dc.contributor.none.fl_str_mv |
Universidade Estadual Paulista (Unesp) |
dc.contributor.author.fl_str_mv |
Marcílio-Jr, Wilson E. [UNESP] Eler, Danilo M. [UNESP] |
dc.subject.por.fl_str_mv |
Dimensionality reduction Explainability Shapley values Visualization |
topic |
Dimensionality reduction Explainability Shapley values Visualization |
description |
Dimensionality reduction (DR) techniques have been consistently supporting high-dimensional data analysis in various applications. Besides the patterns uncovered by these techniques, the interpretation of DR results based on each feature's contribution to the low-dimensional representation supports new finds through exploratory analysis. Current literature approaches designed to interpret DR techniques do not explain the features’ contributions well since they focus only on the low-dimensional representation or do not consider the relationship among features. This paper presents ClusterShapley to address these problems, using Shapley values to generate explanations of dimensionality reduction techniques and interpret these algorithms using a cluster-oriented analysis. ClusterShapley explains the formation of clusters and the meaning of their relationship, which is useful for exploratory data analysis in various domains. We propose novel visualization techniques to guide the interpretation of features’ contributions on clustering formation and validate our methodology through case studies of publicly available datasets. The results demonstrate our approach's interpretability and analysis power to generate insights about pathologies and patients in different conditions using DR results. |
publishDate |
2021 |
dc.date.none.fl_str_mv |
2021-06-25T10:29:21Z 2021-06-25T10:29:21Z 2021-09-15 |
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/10.1016/j.eswa.2021.115020 Expert Systems with Applications, v. 178. 0957-4174 http://hdl.handle.net/11449/206269 10.1016/j.eswa.2021.115020 2-s2.0-85105034882 |
url |
http://dx.doi.org/10.1016/j.eswa.2021.115020 http://hdl.handle.net/11449/206269 |
identifier_str_mv |
Expert Systems with Applications, v. 178. 0957-4174 10.1016/j.eswa.2021.115020 2-s2.0-85105034882 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.relation.none.fl_str_mv |
Expert Systems with Applications |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
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_ |
1803649997155074048 |