摘要
针对判别嵌入式聚类算法对高维数据集聚类运行速度慢的问题,提出一种改进的判别嵌入式聚类算法。利用矩阵的QR分解对类间散度矩阵做特征分解,求得数据的变换预处理;再利用最大间距准则对变换预处理数据再次降维,通过降低判别嵌入式聚类算法时间复杂度来提高效率。对比实验结果表明,改进算法受平衡参数λ的影响较小,平均准确度高于判别嵌入式聚类算法和K均值聚类算法,运行效率也优于判别嵌入式聚类算法。
An efficient Discriminant Embedded Clustering algorithm(EDEC)is proposed in this paper for the slow clustering of Discriminative Embedded Clustering(DEC)algorithm for highdimensional data.In this algorithm,QR decomposition is used for the between-class scatter matrix to make a eig-decomposition and to obtain a transformation preprocessing data,then the Maximum Margin Criterion(MMC)is used to reduce the dimension of the transform preprocessing data,the efficiency is therefore improved by reducing the time complexity of the embedded clustering algorithm.Comparative experimental results show that the improved algorithm is less affected by the balance parameters,that the average accuracy is higher than those by the DEC algorithm and KM clustering algorithm,and that the running efficiency is better than that by the DEC algorithm.
出处
《西安邮电大学学报》
2017年第1期34-37,43,共5页
Journal of Xi’an University of Posts and Telecommunications
基金
陕西省自然科学基金资助项目(2014JM8307)
陕西省教育厅科学研究计划资助项目(14JK1661)
关键词
判别嵌入式聚类
数据降维
最大间距准则
QR分解
discriminative embedded clustering
data dimensionality reduction
maximum margin criterion
QR decomposition