Crisp classifiers vs. fuzzy classifiers: A statistical study

Abstract

A study is made of whether there is a significant statistical difference in performance between crisp and fuzzy rule-based classification. To do that, 12 datasets were chosen from the UCI repository that are widely used in the literature, and use was made of four different algorithms for rule induction - two crisp and two fuzzy - to classify them. Then a non-parametric statistical test was used for measuring the significance of the results, which indicated that both paradigms - crisp and fuzzy classification - are not different in the statistical meaning. © Springer-Verlag 2009.

Publication
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
José Luis Jara
José Luis Jara
Associate Professor