Neuro-Fuzzy and Soft Computing - A Computational Approach to Learning and Artificial Intelligence

Abstract

Single approaches to soft computing have many limitations and disadvantages. Neural network modelling poses a challenge of architecture building, whilst fuzzy sets are characterized by problematic membership functions. The use of hybrid methods is, by contrast, a rather promising strategy. This study aims to develop a new prediction methodology by integrating Neuro-Fuzzy Systems (NFS) with a Neuro-Genetic Approach (NGA). Such a design combines the learning capacity of the neural networks and the ability of fuzzy systems to extract the linguistic knowledge. This proposal is expected to predict the suitability of parameters and models with fewer errors and high accuracy. The performance of this system is improved through the use of genetic algorithm for optimizing the neural network parameters such as the learning rate, network impulse, and the number of membership functions for each input variable. The proposed methodology was tested using electromyographic (EMG) data. The results showed high efficiency (92%) of the proposed hybrid technique.