摘要
在多维有序聚类方法的研究中,现有模型缺乏对离群值的处理,未考虑多维有序数据块状离群的特点,导致聚类结果不合理。为此,基于提取共同趋势序列的思路,提出了两阶段多维有序聚类方法。首先运用动态聚类方法在样本维度摒除离群值影响,提取有序数据的共同趋势序列,然后使用共同趋势序列进行一维有序聚类。此方法不仅可以进行稳健的有序聚类,也可以快速有效地提取有序多维数据中的趋势变化信息。模拟与实证结果显示,考虑了共同趋势提取的多维有序聚类方法,能够提高多维有序样本的信息提取能力和抗干扰性。
In multidimensional ordered clustering analysis,the existing models lack the treatment of outliers,and do not consider the characteristics of multi-dimensional ordered data block outliers which leads to unreasonable clustering results.Based on the idea of extracting common trends,a two-stage multi-dimensional ordered clustering method is proposed.Firstly,the dynamic clustering method is used to eliminate the influence of outliers in the sample dimension,and the common trend sequence of ordered data is extracted.Secondly,it applys the common trend to perform one-dimensional ordered clustering.This method can not only perform robust ordered clustering,but also quickly and effectively extract the trend change information in ordered multi-dimensional data.The simulation and empirical results show that the multi-dimensional ordered clustering method considering common trend extraction can improve the information extraction ability and anti-interference of multi-dimensional ordered samples.
作者
何韩吉
邓光明
HE Han-ji;DENG Guang-ming(College of Science,Guilin University of Technology,Guilin 541004,China;Institute of Applied Statistics,Guilin University of Technology,Guilin 541004,China)
出处
《统计与信息论坛》
CSSCI
北大核心
2020年第12期15-20,共6页
Journal of Statistics and Information
基金
国家自然科学基金项目“快速贝叶斯随机波动建模及其在金融市场中的应用研究”(71963008)。
关键词
多维有序聚类
共同趋势
动态聚类
离群值
Multidimensional ordered clustering
common trend
dynamic clustering
outliers
