Research on universal fuzzy neural comprehensive evaluation network

Abstract

It is a difficult problem to make a perfect comprehensive evaluation on the complex systems with subjectivity and fuzziness. To attain an authentic comprehensive evaluation result in reducing evaluation difficulties, a universal fuzzy comprehensive evaluation network is proposed, which integrates the theory of universal logics and neural networks to serve as a basis for constructing a comprehensive evaluation system with objective weights distribution, appropriate membership functions and adaptive fuzzy compound operators. The comprehensive evaluation network has a simple and intuitively understandable structure, and is not only used to tune the weights distribution and membership functions of fuzzy systems, but also used to tune the compound operators continuously. Its learning techniques can automate this process and substantially reduce development time and cost while improving performance. Therefore, it can be used to a wide range of complex comprehensive evaluation problems. Finally, the practical applications indicate the effectiveness of the proposed model. © 2007 IEEE.