改进的基于广度优先搜索的COP-Kmeans算法
首发时间:2015-07-09
摘要:将广度优先搜索BFS应用于COP-Kmeans算法会对相同的约束对产生不同的搜索序列,导致算法的准确率降低。针对这种情况,提出了一种改进的基于BFS的COP-Kmeans算法。算法首先对训练集进行多次聚类,取得聚类结果,然后对聚类结果进行计算,得到各个聚类结果的标准化互信息,根据标准化互信息计算任意两个数据对象的相关性,最终得到各个数据对象的稳定性,将数据对象稳定性作为数据对象的分配次序的参考依据从而提高算法的准确率,最后重新进行聚类,得到最终的聚类结果。实验结果表明,采用改进后的算法比原先算法在准确率上有所提高。
关键词: 广度优先搜索算法 结合限制的k均值算法 标准化互信息 数据对象稳定性 准确性
For information in English, please click here
An improved COP-Kmeans algorithm based on BFS
Abstract: The breadth first search applied to the COP-Kmeans algorithm will produce different search sequences for the same constraints,which will reduce the accuracy of the algorithm . In view of this, we propose an improved COP-Kmeans algorithm based on the breadth first search (BFS). Firstly, we trained the data set to get the clustering results, and then the calculation results are calculated to get the normalized mutual information of each calculation result, according to the normalized mutual information we can calculate the correlation between any two instances, take the instance stability as a reference to distribute the instance will improve the accuracy of the algorithm. Finally re-clustering to get the final clustering result. . Experimental results show that the improved algorithm can obtain an increase in the accuracy compared to the original algorithm.
Keywords: Breadth first search Constrained K-means Clustering Normalized mutual information The instance stability Accuracy
基金:
论文图表:
引用
No.4647089107235214****
同行评议
共计0人参与
勘误表
改进的基于广度优先搜索的COP-Kmeans算法
评论
全部评论0/1000