摘要
CLARANS算法是一种有效且广泛应用的聚类算法,适合发现任意形状的聚类结果,但CLARANS算法在搜索过程中容易陷入局部最优解,从而忽略全局最优解。为了避免CLARANS算法在搜索中心点时易受局部最优解的影响,提出一种将CLARANS算法中的邻接点作为QPSO算法的量子粒子,结点代价作为适应度函数对其进行寻优的改进CLARANS算法。将该改进算法应用于UCI数据集,结果表明该算法聚类效果好、收敛快,算法的稳定性、收敛性及寻优能力都有很大提高。
CLARANS algorithm is an efficient and effective and wide application clustering algorithm. It is applicable to locate objects with polygon shape. CLARANS often gets stuck at a locally optimum configuration, ignores the global optimum solution. This paper presents an improved CLARANS algorithm based on the QPSO algorithm in order to avoid local optimum. The improved method adopts the quantum particle as the neighbor and takes the node cost as the fitness function. The improved CLARANS algorithm is applied to the UCI data set. The simulation experiment results show that it can improve the clustering performance.
出处
《计算机工程与应用》
CSCD
2013年第9期168-170,179,共4页
Computer Engineering and Applications
基金
湖南省教育厅基金资助项目(No.11C1025)
关键词
量子粒子群算法
基于随机选择的聚类算法(CLARANS)算法
结点代价
聚类
适应度函数
Quantum Particle Swarm Optimization （QPSO） algorithm
Clustering Algorithm based on Randomized Search（CLARANS） algorithm
node cost
clustering
fitness function
