Explaining dimensionality reduction results using Shapley values

dc.contributor.authorMarcílio-Jr, Wilson E. [UNESP]
dc.contributor.authorEler, Danilo M. [UNESP]
dc.contributor.institutionUniversidade Estadual Paulista (Unesp)
dc.date.accessioned2021-06-25T10:29:21Z
dc.date.available2021-06-25T10:29:21Z
dc.date.issued2021-09-15
dc.description.abstractDimensionality 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.en
dc.description.affiliationFaculty of Sciences and Technology São Paulo State University (UNESP)
dc.description.affiliationUnespFaculty of Sciences and Technology São Paulo State University (UNESP)
dc.identifierhttp://dx.doi.org/10.1016/j.eswa.2021.115020
dc.identifier.citationExpert Systems with Applications, v. 178.
dc.identifier.doi10.1016/j.eswa.2021.115020
dc.identifier.issn0957-4174
dc.identifier.scopus2-s2.0-85105034882
dc.identifier.urihttp://hdl.handle.net/11449/206269
dc.language.isoeng
dc.relation.ispartofExpert Systems with Applications
dc.sourceScopus
dc.subjectDimensionality reduction
dc.subjectExplainability
dc.subjectShapley values
dc.subjectVisualization
dc.titleExplaining dimensionality reduction results using Shapley valuesen
dc.typeArtigo

Arquivos

Coleções