面板数据加权聚类分析方法研究

在面板数据聚类分析方法的研究中,基于面板数据兼具截面维度和时间维度的特征,对欧氏距离函数进行了改进,在聚类过程中考虑指标权重与时间权重,提出了适用于面板数据聚类分析的"加权距离函数"以及相应的Ward.D聚类方法。首先定义了考虑指标绝对值、邻近时点增长率以及波动变异程度的欧氏距离函数;然后,将指标权重与时间权重通过线性模型集结成综合加权距离,最终实现面板数据的加权聚类过程。实证分析结果显示,考虑指标权重与时间权重的面板数据加权聚类分析方法具有更好的分辨能力,能提高样本聚类的准确性。

Study on Weighted Clustering Analysis Method with Panel Data

In the study of clustering analysis method with panel data,the Euclidean distance function improved based on the time dimension feature and the indicators dimension feature of the panel data.Considering weights of index and time in the process of clustering,this paper put forward a weighted distance function that is suitable for clustering analysis in panel data and a corresponding clustering method-Ward.D.Firstly,define the indexes of absolute value,adjacent point of growth,as well as the volatility variation of Euclidean distance function.Then,construct a comprehensive weighted distance by the linear model with the time weight and the index weight.Finally,realize the weighted clustering analysis in panel data.The empirical analysis shows that considering the weight of index and time about the weighted clustering analysis method on panel data has better resolution,which can improve the accuracy of the cluster about the sample.