改进的量子遗传进化激励的快速数据分类算法  

Improved Quantum Genetic Evolution Incentive Fast Clustering Algorithm

作　　者：代亮[1] 谢晓尧[2] 

机构地区：[1]贵州师范大学数学与计算机科学学院,贵州贵阳550001 [2]贵州师范大学重点实验室,贵州贵阳550001

出　　处：《计算机仿真》2014年第2期340-343,共4页Computer Simulation

基　　金：黔科合J字LKS[2012]34号

摘　　要：由于数据的属性越来越复杂,可用于分类的属性越来越多。传统的数据聚类算法对数据聚类初始中心点十分敏感、很难根据数据复杂的属性选定合适的中心,计算过程中由于数据的多属性干扰易陷入计算的局部最优,造成算法执行效率较低。提出了一种改进的量子遗传进化激励的快速聚类算法。算法中,首先利用高密度划分和阈值参数对样本数据集进行初次聚类划分,生成若干聚类集合;然后将样本的聚类过程看成是聚类中心的动态优化过程,克服传统算法中中心点固定不变的弊端,利用改进的量子遗传进化激励算法对每次聚类最优的聚类中心进行搜索;算法引入自适应变异算子对进化算法进行搜索能力进行改进,增强算法的全局搜索能力。实验证明,算法不仅具有较好的聚类精度,而且收敛速度快。Since the properties of data is complex, the properties of data to be classified are increased, the traditional clustering algorithms are very sensitive to initial focal point for data clustering, and it is difficult to select ap- propriate center according to the complexity characteristic of the data. This paper presented an improved fast cluste- ring algorithm based on quantum genetic evolution incentive. In this algorithm, we used high - density partition and threshold parameters for the initial set of the sample data clusters to generate a number of clusters set. Finally, the optimal cluster center of each cluster was searched by using the improved algorithm. The experimental results show that the algorithm introduces adaptive mutation operator of evolutionary algorithms to improve search capabilities, enhances global search capability. Furthelmore, the clustering algorithm has a better accuracy and fast convergence speed.

关 键 词：聚类 遗传算法 粒子群 密度聚类 激励 

分 类 号：TP311.4[自动化与计算机技术—计算机软件与理论]