论文部分内容阅读
以改进的流形距离为相似度测度,结合人工蜂群算法,提出一种二阶段聚类算法.首先根据局部密度、最大最小距离和近邻选择对数据集初步归类并得到簇代表点;然后将聚类归属为优化问题,通过改进的蜂群算法对簇代表点及没归类的样本点较快地搜索到最优聚类中心,同时根据流形距离的全局一致性特征,对样本进行精确的类别划分;最后将两阶段算法综合归类.实验结果表明,所提出的算法可以获得良好的聚类效果.
This paper proposes a two-stage clustering algorithm based on improved manifold distance and artificial bee colony algorithm. Firstly, the data set is initially classified according to the local density, maximum and minimum distance and nearest neighbor selection, and then the cluster representative points are obtained. Then The clustering is attributed to the optimization problem. The improved clustering algorithm is used to search the cluster centers and the uncategorized samples quickly for the optimal cluster centers. At the same time, according to the global consistency of manifold distances, Accurate category classification.At last, the two-stage algorithm is comprehensively classified.The experimental results show that the proposed algorithm can get a good clustering effect.