Fuzzy relation-based polynomial neural networks based on hybrid optimization

Abstract

This paper introduces hybrid optimized fuzzy relation-based polynomial neural network (HOFRPNN), a novel architecture that is constructed by using a combination of fuzzy rule-based models, polynomial neural networks (PNNs) and a hybrid optimization algorithm. The proposed hybrid optimization algorithm is developed by a combination of a space search algorithm and an improved complex method. The structure of HOFRPNN comprises of a synergistic usage of fuzzy-rule-based polynomial neuron that are essentially fuzzy rule-based models and polynomial neural networks that is an extended group method of data handling (GMDH). The architecture of HOFRPNN is an essentially modified PNN whose basic nodes are fuzzy-rule-based polynomial neurons rather than conventional polynomial neurons. Moreover, the hybrid optimization algorithm is utilized to optimize the structure topology of HOFRPNN. A comparative study demonstrates that the proposed model exhibits higher accuracy and superb predictive capability when compared with some previous models reported in the literature. © 2012 Springer-Verlag.