Fuzzy kernel self-organizing map neural network and the application

The fuzzy kernel self-organizing map neural network is presented and a corresponding learning algorithm is proposed to train this network. There are two mainly improvement points to be shown through comparing the algorithm with the basic SOM algorithm. The first point, the kernel method is directly introduced to extend the SOM algorithm so that the extended algorithm can deal with nonlinear data in original space. The second point, through combining both the fuzzy theory and self-organizing map neural network, topology preservation is enhanced. Two simulated researches of the proposed algorithm in classifying data of classical datasets and acrylonitrile reactor is provided. The experimental results show that this algorithm can obtain better clustering results and clustering results can be used to monitor yields and guide to optimize operational parameters.