An Extended Compensated Fuzzy C-Means Algorithm and Its Application in Optimizing the Acrylonitrile Reactor Parameters

An extended compensated fuzzy c-means (ECFCM) algorithm is proposed. Considering the relation that actually exists between input training data and clustering centers, grey relational coefficient is introduced to extend compensated fuzzy c-means (CFCM) algorithm. In addition, for expanding the applications of the proposed method, the function of clusters´ integration and division is added into its run process so as to automatically obtain the rational number of clusters. Our simulated experiments demonstrate that the results of the proposed method applied to analysis of the iris data and the wine data outperform those of other methods. Furthermore, a practical application in classifying the acrylonitrile reactor data is also provided. The experimental results show that the ECFCM algorithm with the function of classes´ integration and division can obtain the rational number of clusters and better clustering performance. Higher yields can be obtained by regulating the acrylonitrile reactor parameters according to the clustering results. Therefore, the ECFCM algorithm can be well applied to optimizing the acrylonitrile reactor parameters