摘要
聚类融合通过把具有一定差异性的聚类成员进行组合,能够得到比单一算法更为优越的结果,是近年来聚类算法研究领域的热点问题之一。提出了一种基于自适应最近邻的聚类融合算法ANNCE,能够根据数据分布密度的不同,为每一个数据点自动选择合适的最近邻选取范围。该算法与已有的基于KNN的算法相比,不仅解决了KNN算法中存在的过多参数需要实验确定的问题,还进一步提高了聚类效果。
The clustering ensemble algorithms can get a more superior result than the single clustering algorithms, because clustering ensemble combines clustering members which have some difference in each other. The problem of clustering ensemble algorithms has become one of the hot field in recent years. Based on the concept of adaptive nearest neighbors, this paper proposes a new clustering ensemble Adaptive Nearest Neighbors Clustering Ensemble （ANNCE） algorithm. This algorithm can automatically select a different number of nearest neighbors for each data point, according to the different density of data, and compared with the algorithm based on KNN, the ANNCE algo- rithm can not only solve the problem of too many parameters but also get a better result.
出处
《计算机工程与应用》
CSCD
2012年第19期157-162,共6页
Computer Engineering and Applications
基金
国家科技支撑计划项目(No.2009BAH42B02)
国家自然科学基金(No.60873038)
关键词
聚类融合
自适应最近邻
ANNCE算法
clustering ensemble
adaptive nearest neighbors
ANNCE algorithm
