A fuzzy neural network for knowledge acquisition in complex time series

A novel fuzzy neural network, called FuNN, is applied here for time-series modelling. FuNN models have several features that make them well suited to a wide range of knowledge engineering applications. These strengths include fast anti accurate learning, good generalisation capabilities, excellent e xplanation facilities in the form of semantically:ally meaningful fuzzy rul es, and the ability to accommodate both numerical data and existing expert knowledge about the problem under consideration. We investigate the effecti veness of the proposed neuro-fuzzy hybrid architectures for manipulating th e future behaviour of nonlinear dynamical systems and interpreting fuzzy if -then rules. A well-known example of Box anti Jenkins is used as a benchmar k time series in the proposed modelling approach and the other modelling ap proach. Finally, experimental results and comparisons with the other popula r neuro-fuzzy inference system, namely Adaptive Network-based Fuzzy Inferen ce System (ANFIS) are also presented.