摘要
针对传统K-means算法存在的缺陷,引进人工鱼群算法,提出了一种基于改进鱼群和K-means的混合聚类算法。聚类样本中心点初始化时,人工鱼各维参数随机选择在对应属性两个极值之间,同时为了降低计算复杂度,提高收敛效率,寻找全局最优,首先对随机选取的一小部分人工鱼进行K-means操作,然后对全体人工鱼的追尾算子引入粒子群策略,引导其学习,模拟人工鱼的行为。通过Matlab仿真实现算法,在费雪鸢尾花卉数据集和葡萄酒质量数据集进行了实验,算法的有效性和可行性得到了验证。
In order to overcome the existing shortcoming of traditional k-means clustering algorithm, this paper introduces Artificial Fish Swarm Algorithm (AFSA). A new hybridized algorithm is proposed for data clustering based on improved artificial fish swarm algorithm and k-means algorithm. Randomly select initial center pointer between the two extremes about attributes, in order to reduce the computational complexity, improve the convergence efficiency, find the global optimum, performed k-means on some artificial fishes randomly, integrated particle swarm strategy into the follow operator to guide the learning of artificial fishes, simulate the behaviors of artificial fishes. Achieve this integrated algorithm in Matlab, experiment on the Iris datasets and wine datasets, the effectiveness and feasibility of the algorithm has been verified.
出处
《计算机工程与应用》
CSCD
2013年第22期119-122,共4页
Computer Engineering and Applications
基金
国家社会科学基金(No.08CTQ014)
大学数字图书馆国际合作计划(No.B2014)
关键词
人工鱼群
K-均值
聚类
粒子群
混合算法
Artificial Fish Swarm Algorithm (AFSA)
k-means
data clustering
Particle Swarm 0ptimization(PSO)
hybrid algorithm