基于选择和变异机制的蛙跳FCM算法

为了改进模糊C-均值(FCM)聚类算法对初始值和噪声数据敏感,且易陷入局部极小值的缺点,提出一种基于选择和变异机制的蛙跳FCM算法(SMSFLA-FCM)。该算法首先将线性递减的惯性权重引入蛙跳算法的更新策略中,按照一定的概率选择适应度值较优的青蛙代替较差青蛙,并对每只青蛙个体以不同的概率变异;再用改进后的蛙跳算法求得最优解作为FCM算法的初始聚类中心;然后利用FCM优化初始聚类中心;最后求得全局最优解,从而有效克服了FCM算法的缺点。人造数据和经典数据集的实验结果表明,SMSFLA-FCM与SF-LA-FCM和FCM聚类算法相比,提高了算法的寻优能力,且迭代次数更少,聚类效果更好。

Fuzzy clustering algorithm based on selection and mutation mechanism shuffled frog leaping algorithm

Because of the problems such as the sensitivity to initial value and noise data with the fuzzy C-means(FCM)clustering algorithm and its ready occurrence of local minimum,this paper presented a fuzzy C-means clustering based on selection and mutation mechanism shuffled frog leaping algorithm.This algorithm introduced the linear decreasing inertia weight to correct the poor frog update strategy.Then it selected the frog with better fitness value to substitute the poor one,and made very frog to mutate with different probability.The optimal solution obtained with SMSFLA with strong global searching ability was taken as initial clustering centers of FCM algorithm to optimize initial clustering centers,so as to get the global optimum and overcome the shortcoming of the FCM algorithm.The results of experiment on the artificial and real data show that compared with the SFLA-FCM and FCM clustering algorithm,the new algorithm(SMSFLA-FCM)optimization ability would be stronger,the number of iterations less,and the clustering efficiency better.