Monotonicity aspects of linguistic fuzzy models

Their interpretable model structure sets linguistic fuzzy m models apart from other modelling techniques and is considered their greatest asset. Therefore, in the identification process of a linguistic fuzzy model, the interpretability of the model should be safeguarded or at least be balanced against its accuracy. A good trade-off between accuracy and interpretability can be obtained by including as much qualitative knowledge as possible in the data-driven model identification process. Monotonicity is the type of qualitative knowledge that plays a central role in this dissertation. Monotone is hereby interpreted as order-preserving. This dissertation contributes to the ecological modelling domain by the application of fuzzy ordered classifiers to a habitat suitability modelling problem of river sites along springs to small rivers in the Central and Western Plains of Europe for 86 macroinvertebrate species. Furthermore, it contributes to the fuzzy modelling domain by (1) introducing an accurate and fast computational method for determining the crisp output of Mamdani-Assilian models applying the Center of Gravity defuzzification method and using fuzzy output partitions of trapezial membership functions, (2) presenting a new performance measure for fuzzy ordered classifiers, referred to as the average deviation (AD) as it takes the ordering of the output classes into account, (3) formulating guidelines for designers of monotone linguistic fuzzy models and (4) introducing a new inference procedure, called ATL-ATM inference, for linguistic fuzzy models with a monotone rule base.

Author dc:creator

• Van Broekhoven, Ester

Contributors dc:contributor

• De Baets, B

Statement dc:rights

• info:eu-repo/semantics/openAccess

Language dc:language

und

Identifier

https://biblio.ugent.be/publication/469506
http://doi.org/1854/7200
https://biblio.ugent.be/publication/469506/file/1879580

OAI identifier oai:identifier

oai:archive.ugent.be:469506