Study On The Panel Data Clustering Analysis Method And Empirical Analysis

As an important analytical method of multivariate statistical analysis,cluster analysis has been widely used in many fields.But the subject of traditional clustering method is usually the cross section data on a fixed time,these methods cannot be directly applied to the panel data clustering,so it is necessary to find suitable clustering method for panel data.In the case of cluster analysis,the similarity measurement between the clustering objects needs to be first defined regardless of the clustering technique.Because panel data has two aspects of time dimension and index dimension,how to make full use of the information of these two aspects is the key when defining the similarity measure of panel data.This paper,on the basis of existing research,studied the clustering method of panel data.The main research contents are as follows:1.This paper proposed a clustering analysis method based on triangular grids.Through triangulation of panel data surfaces,analyzing its geometry as characterization of individual development trend.Using an exponential regulation function to integrate the distance between the individual and the similarity of shape,considering its cross-section static information and dynamic information changing over time.The new distance based on the exponential adjustment function is defined as the measure standard of the similarity between individuals.On this basis,the traditional clustering method is used to cluster analysis.2.Through the clustering method based on the triangular mesh proposed in this paper,the air quality condition of 18 new first-tier cities in China is analyzed.The validity of clustering method is verified by variance analysis.Finally,these 18 cities are divided into five categories.The characteristics and evolution of each kind of city are analyzed,studying the difference between each type of city.3.We compared the clustering results with the clustering results of square Euclidean distance to analyze the impact of exponential adjustment function on the distance between the individuals,and comparing with the cross section data clustering results for dynamic analysis.Through the comparison between clustering results,it can be found the clustering result of the method proposed in this paper,can reflect the differences on the index dimension between the individual and the comprehensive change on the time dimension.Therefore,the clustering method based on triangular meshing is suitable for panel data clustering analysis.