摘要
针对模糊聚类算法对初值和聚类中心较为敏感的问题,采用和声搜索算法寻找最优聚类中心,并且改进了和声搜索算法的调音概率和随机带宽,从而加速了算法收敛。使用维度加权的方法进行特征选择,提高了聚类的性能,通过定义聚类质量评价函数提高了模糊聚类质量。采用标准数据验证了算法。结果表明,提出的聚类算法性能优于其他同类算法。
As fuzzy clustering algorithm is more sensitive to initial values and cluster centers, the paper presents a modified harmony to rapidly and efficiently find the optimal cluster centers, which adjusts pitch rate and random bandwidth of original harmony search algorithm to accelerate convergence rate. The weighted dimension is used to select feature in order to improve the clustering performance and the clustering quality evaluation function is defined to improve the clustering quality. Finally, the paper uses standard dataset to validate some algorithms and the results show that the proposed clustering algorithm outperforms other similar algorithms.
出处
《重庆理工大学学报(自然科学)》
CAS
2012年第8期71-78,共8页
Journal of Chongqing University of Technology:Natural Science
基金
教育部基金资助项目(10YJC870037)
重庆市教委科学研究项目(KJ100805)
关键词
和声搜索
维度加权
特征选择
聚类
harmony search
weighted dimension
feature selection
clustering