一种基于密度函数的直觉模糊聚类初始化方法

Initialization Method for Intuitionistic Fuzzy Clustering Based on Density Function

针对基于目标函数的直觉模糊聚类方法容易陷于局部最优值的问题,提出了一种改进的密度函数初始化方法。该方法首先利用样本密度函数在较高局部密度的区域中选取c个样本,然后遍历剩余样本进行粗归类,并计算每类各维数据的平均值作为初始聚类中心。最后通过典型实例验证,该方法不仅解决了容易陷入局部极小值的问题,同时迭代次数减少,收敛速度加快,提高了聚类性能。

To the problems of the technique of Intuitionistic fuzzy clustering based on objective function immersed in partial optimization figure, an improved initialization method based on density function was proposed. First, the proposed method have selected e samples in the upper partial density making use of sample density function. Then the other samples were visited and classified roughly. Moreover, the every kind of average values of all dimensions were calculated, which are counted as initial clustering center. At last, the validity of the technique proposed was checked with an classical instance,and the technique not only solved the problem of functions easily immersed in partial optimization figure, which takes on, but also reduced iterative times and improve convergent speed, so as to improve clustering capability.