CF-integrals: A new family of pre-aggregation functions with application to fuzzy rule-based classification systems

0485363 - ÚTIA 2019 RIV US eng J - Článek v odborném periodiku
Lucca, G. - Sanz, A. - Dimuro, G. P. - Bedregal, B. - Bustince, H. - Mesiar, Radko
CF-integrals: A new family of pre-aggregation functions with application to fuzzy rule-based classification systems.
Information Sciences. Roč. 435, č. 1 (2018), s. 94-110. ISSN 0020-0255. E-ISSN 1872-6291
Institucionální podpora: RVO:67985556
Klíčová slova: CF-integral * classification problems * pre-aggregation
Obor OECD: Pure mathematics
Impakt faktor: 5.524, rok: 2018
http://library.utia.cas.cz/separaty/2018/E/mesiar-0485363.pdf

This paper introduces the family of CF-integrals, which are pre-aggregations functions that generalizes the Choquet integral considering a bivariate function F that is left 0-absorbent. We show that CF-integrals are 1->-pre-aggregation functions, studying in which conditions they are idempotent and/or averaging functions. This characterization is an important issue of our approach, since we apply these functions in the Fuzzy Reasoning Method (FRM) of a fuzzy rule-based classification system and, in the literature, it is possible to observe that non-averaging aggregation functions usually provide better results. We carry out a study with several subfamilies of CF-integrals having averaging or non-averaging characteristics. As expected, the proposed non-averaging CF-integrals obtain more accurate results than the averaging ones, thus, offering new possibilities for aggregating accurately the information in the FRM. Furthermore, it allows us to enhance the results of classical FRMs like the winning rule and the additive combination.
Trvalý link: http://hdl.handle.net/11104/0281414