基于多维面板数据的聚类方法探析及实证研究

面板数据由于能够从截面和时间构成的二维空间来描述研究对象的动态特征而被广泛应用于经济问题的建模实践中。本文借鉴多元统计学中主成分分析方法对面板数据进行降维处理,然后通过构建综合评价函数序列矩阵的相似指标对面板数据进行聚类分析,并提出一些研究面板数据亲疏关系的有效途径,最后运用该算法对我国地区科技能力进行实证分析,结果与实际状况较为吻合.

The Discussion on the Clustering Way Based on the Multi-dimensional Panel Data and Empirical Analysis

Panel data is widely used in modeling on economic problems because it can describe the object's dynamic characters from time and cross-section in two-dimensional space.This paper reduces dimension of panel data by the means of principal component analysis,then uses cluster analysis on the panel data through similarity indexes which describe the comprehensive sequential evaluation matrix, lastly puts forward the effective research approaches on the panel data's affinities.After that the authors use this algorithm to analyze the regional capacity of science and technology in China,analysis result comparatively tallies with actual situation.