Objective Measures Ensemble in Associative Classifiers

• Autor(a) principal: Dall'Agnol, Maicon [UNESP]
• Data de Publicação: 2020
• Outros Autores: Carvalho, Veronica Oliveira de [UNESP], Filipe, J., Smialek, M., Brodsky, A., Hammoudi, S.
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.5220/0009321600830090; http://hdl.handle.net/11449/210700
• topic: Associative Classifier; Interestingness Measures; Ranking; Classification; Association Rules

Associative classifiers (ACs) are predictive models built based on association rules (ARs). Model construction occurs in steps, one of them aimed at sorting and pruning a set of rules. Regarding ordering, usually objective measures (OMs) are used to rank the rules. The aim of this work is exactly sorting. In the proposals found in the literature, the OMs are generally explored separately. The only work that explores the aggregation of measures in the context of ACs is (Silva and Carvalho, 2018), where multiple OMs are considered at the same time. To do so, (Silva and Carvalho, 2018) use the aggregation solution proposed by (Bouker et al., 2014). However, although there are many works in the context of ARs that investigate the aggregate use of OMs, all of them have some bias. Thus, this work aims to evaluate the aggregation of measures in the context of ACs considering another perspective, that of an ensemble of classifiers.