摘要
针对运用半监督仿射传播聚类算法处理高维数据时聚类精度低和计算量大的问题,提出一种基于局部线性嵌入的半监督仿射传播聚类算法.该算法首先通过LLE算法将高维输入数据集映射到低维空间得到低维数据集,计算低维数据集的相似度矩阵,再用半监督算法调整相似度矩阵,最后用仿射传播聚类算法对低维数据进行聚类分析.仿真结果表明,本文提出的算法与半监督仿射传播聚类算法相比,在处理高维数据时聚类效果更好,精度更高,迭代次数更少.
Aimed at the problem of low clustering precision and time-consuming calculation of high-dimensional data sets with semi-supervised affinity propagation clustering algorithm,a local linear embeddingbased clustering algorithm of semi-supervised affinity propagation（LLE-SAP）is proposed.Firstly,with this algorithm,the high-dimensional input data set will be mapped into low-dimensional space with LLE algorithm to get a low-dimensional data set.Then the similarity matrix of the low-dimensional data is evaluated and adjusted with semi-supervised algorithm.Finally,the clustering analysis of the low-dimensional data will be conducted with affinity propagation clustering algorithm.Simulation result shows that compared with the algorithm available,the proposed algorithm will have higher precision,fewer iterations,and better effect for treating high-dimensional data.
出处
《兰州理工大学学报》
CAS
北大核心
2015年第1期96-100,共5页
Journal of Lanzhou University of Technology
基金
国家自然科学基金(51265032
61263003)
甘肃省高校基本科研业务费项目(1203ZTC061)
关键词
数据挖掘
半监督
仿射传播聚类
局部线性嵌入算法
data mining
semi-supervision
affinity propagation clustering
local linear embedding algorithm
