摘要
针对基于进化算法的自动聚类方法具有收敛速度慢的缺陷,为回忆收敛性,提高算法精度,提出一种改进的差分进化自动聚类算法。算法从改进染色体评价过程中的解码方式,依据由染色体解码得到的聚类数和质心集,通过质心筛选和质心聚类两步操作,从包含于染色体中的聚类划分簇中提取较优的聚类划分,从而避免了因随机解码方法导致的对染色体的错误评价,使较优的染色体能够在种群进化中存活下来。仿真结果表明,新算法的收敛速度明显好于同类算法,并且收敛精度也有改善。
Automatic clustering algorithms based on evolution computation have slow convergence rates.In this paper,we propose a modified algorithm with faster convergence rate based on differential evolution.The departure point of the proposed algorithm is to improve the way chromosomes are rated in the decoding process.Based on the number of clusters and the set of centroids obtained from chromosome decoding,the algorithm can obtain an optimized partitioning of the data set through the operations of selecting and clustering,and can thus effectively avoid incorrect ratings for the chromosomes due to random decoding.In this way,the better chromosomes are preserved in the evolution process.Simulation results show that the new algorithm can achieve much faster convergence rate and improved the accuracy than alternative methods.
出处
《计算机仿真》
CSCD
北大核心
2010年第11期69-72,135,共5页
Computer Simulation
关键词
自动聚类
差分进化
全局优化
Automatic clustering
Differential evolution（DE）
Global optimization
