直觉模糊集合数据的聚类有效性分析

Clustering Validity Analysis for Data of Intuitionistic Fuzzy Sets

针对直觉模糊集合数据的聚类有效性问题,给出了一种用于发现最优模糊划分的聚类有效性方法。该方法采用直觉模糊相关度和直觉模糊熵两个重要因子来评价直觉模糊聚类的有效性。其中,直觉模糊相关度通过增加非隶属度参数对模糊相关度进行直觉化扩展,用于评价类与类间相关度的大小,同时加入权重参数解决了样本数据各维特征分配不均匀的问题,而直觉模糊熵用于检验分类结果的可靠性。最后通过实例验证了该方法对于紧致的、良好分离的数据集分类效果理想,其在目标编群、目标识别等信息融合领域有良好的应用前景。

To measure clustering validity for the data of intuitionistic fuzzy sets, a clustering validity method is proposed to identify the optimal fuzzy partition. The method exploits two important evaluation factors： intuitionistic fuzzy correlation and intuitionistic fuzzy entropy. By adding the non-membership degree parameter to broaden intuitively fuzzy correlation, the first factor can evaluate correlation values during classes. Meanwhile, it solves the training data of problem with asymmetrical weight on every dimension character by adding the weighted parameters. Furthermore, the second factor is used to verify the reliability of clustering results. Finally, experimental results show that the method has the favorable effect on tighten and well-detached data sorts, and preferable application prospect in information fusion field, such as target classification and target recognition.