Fuzzy discrimination analysis method based on RBFNN and its application in soft measurement

Artificial neural networks(ANN) has being used widely in information processing, intelligence control because of its abilities of self-organization, self-learning and parallel-processing. The work to use ANN theory and Fuzzy sets together to solve the practical problem is being promoted with the fuzzy sets birth and development. Based on the basic theory of RBFNN and fuzzy sets, a new fuzzy discrimination analysis method named RFD(fuzzy discrimination based on RBFNN) is proposed in this paper. It includes three steps to construct RFD, that is classification earmark, modeling by RBFNN and fuzzy reasoning. After explaining in detail the process of the above three steps to construct RFD. It is tested for discriminating some UCIs such as the IRIS data and a practical soft measurement data. The results indicate that RFD has low error- discrimination percentage, short discrimination time and satisfied practical application effect.