Supervised fuzzy clustering for rule extraction

Self-spawning neuro-fuzzy system for rule extraction

“…Building a meaningful fuzzy inference system to a large extent is to establish a series of fuzzy rules that cover all the rule patches in the input space. At the same time, we must keep the number of rules as low as possible in order to maintain the generalizing ability of the model and to ensure a compact and transparent model (Setnes 2000). Therefore, our task is reduced to seeking an optimal number of rule patches, locating their positions, and computing their shape widths in the input space.…”

“…Therefore, it is important to learn knowledge and derive fuzzy rules from the data itself (Cherkassky and Mulier 1998). Various methods have been proposed, e.g., fuzzy clustering in product space (Setnes 2000) or in augmented data set , rule generation using data condensation algorithm (Mitra et al 2002), genetic algorithms (Lee and Takagi 1993;Ishibuchi et al 1995), entropy (Yager and Filev 1993), orthogonal transformation methods (Wang and Mendel 1992;Yen and Wang 1999), and a group of neuro-fuzzy approaches (Berenji and Khedkar 1992;Brown and Harris 1994;Gupta and Rao 1994;Jang 1993;Fritzke 1997), etc.…”

“…It is desirable that the rule base covers all the states of the system that are important to data understanding (Duch et al 2001;Duch et al 2004) and the intended application (Setnes 2000). In general, fuzzy rules can be obtained by either human experts or a data-driven extraction scheme from measured input-output data pairs.…”

Noninvasive data analysis: a web of information granules

“…[4,7,9]. In what follows, we are interested in the notion of membership cast in the framework of pattern recognition.…”

“…The data mining trend is augmented by a general methodology of Computational Intelligence (CI) and in this way is linked to fuzzy modeling, cf. [5,6,9,10]. Neural networks being an integral component of CI contribute to the flexibility (adaptive properties) of the systems.…”