Neuro-fuzzy system for large data sets

Abstract

The paper describes the neuro-fuzzy system for large data sets. The large data set is split into subsets and independent submodels are elaborated. The models are then merged. The described approach enables realisation of incremental learning paradigm. The paper proposes new measure of rule quality based on the logical implications and measure for similarity of rules in neuro-fuzzy systems. The theory is accompanied by experimental results. © 2011 Springer-Verlag Berlin Heidelberg.