模糊C-均值(FCM)聚类算法的实现

传统的FCM算法能够将靠近边界的具有固有形状的两个簇合并成为一个大的簇.然而,对于一些稍微复杂的数据,如果没有其它的像去除小簇之类的机制的话,FCM算法很难将非常接近的类聚类到一起.给出的聚类算法是在传统FCM算法的循环之后添加了去除掉空簇的步骤,解决了上述很难将非常接近的类聚到一个簇中的问题.另外,为便于选出最优结果,在递归之后又添加了计算聚类有效性的步骤.最后用Java实现了该算法并在数据集上进行了实验,证实了改进方法的有效性.

THE IMPLEMENTATION OF THE FUZZY C-MEANS CLUSTERING ALGORITHM

The traditional FCM algorithm lumps the two clusters close to boundaries with natural shapes into a large cluster. However,for some complex data, it is hard for the FCM to cluster the very close classes together without the help of other mechanisms such as mechanism for elimination of small clusters. The step of eliminating empty clusters is added after the loop of traditional FCM ,and the clustering problem is solved. In addition, in order to choose the optimum result, a step of computing clustering validity is added after the iteration of FCM. Finally, an implementation of this algorithm in java is given, and the test on the dataset proves the validity of the algorithm.