一种改进型专家模糊核聚类赋权方法研究

针对传统基于判断矩阵的专家模糊核聚类赋权方法，由于归一化条件的制约，导致离群点对聚类结果产生不良影响的问题，提出一种改进型模糊核聚类算法。该方法在聚类过程中，通过放宽归一化约束条件，削弱离群点对聚类结果的影响；并且针对传统基于信息熵与一致性系数线性耦合的聚类标准的局限性，提出一种基于偏差熵的赋权方法，依据专家对自身类别的聚类贡献度，确定专家权重，克服了传统方法的不足。算例表明，该方法可行、有效。

Research on an Improved Expert Cluster Weighting Method Based on Fuzzy Kernel Clustering

In order to settle the matter that the outliers have an adverse impact on clustering results, which is caused by the restriction of traditional normalization conditions, an improved expert cluster weighting method based on fuzzy kernel clustering is proposed. By loosening the normalization constraints, the effects of outliers on the clustering results could be overcome. At the same time, considering the limitation of the traditional clustering standard based on the linear coupling of information entropy and consistency coefficient, a cluster weighting method based on deviation entropy is proposed. The expert weights' is determined based on the contribution of the expert clustering, which overcomes the deficiency of previous method. The example shows that the method is feasible and effective.