创意FCM算法

针对现有模糊聚类方法仅仅是对已有数据点的聚类的不足,提出了在已有数据集的基础上找到新的一类集群的聚类方法 CFCM。该算法在FCM算法的基础上,通过引入观测点P作为聚类的先验知识,来大致确定未知集群的聚类中心,定义了权重系数λ来限定观测点对新的一类聚类中心形成的影响程度。人造数据集和UCI真实数据集的实验结果表明,该算法不仅对已知数据点有较好的聚类效果,并且可以在观测点P的作用下在特定区域创造出新的一类无已知数据点的集群中心点的大致位置,因而在实际中有潜在应用价值。

A creative fuzzy c-means clustering algorithm

Given that the existing fuzzy clustering methods are just for existing data points, we put forward a new clustering method based on the original dataset to find a new cluster, called creative fuzzy c-means clustering algorithm (CFCM). The algorithm based on the FCM algorithm roughly determines the clustering center of the unknown cluster by introducing the observation point P as the prior knowledge of clustering, and defines the weight coefficient λ to confine the influence degree of the point P on the formation of the new clustering center. Experimental results on artificial datasets and UCI classic datasets show that the proposed clustering algorithm not only has good clustering effect for the known data points, but can also creatively find out the new clustering center for certain zone which is only indicated by an observation point and does not contain any known data points, thus having potential applications in practice.