Research On The Models And Application Based On Grey Relational Analysis And Clustering To The Panel Data

For the problem of Grey Relational Analysis and Correlation Clustering for the panel data,basedon the classic Grey Relational Analysis and Grey Clustering methods,the Grey Relational Analysismodel based on the Principal Component Analysis and Grey Correlation Clustering model for thepanel data are proposed,to improve the Grey Relational Analysis method to solve the problem of thecorrelation analysis and the Grey Clustering analysis to the small samlpe panel data.The specificresearch is showed in the following:(1) Given that the sequences of system behavior and related factors described by the form of thepanel data,the Grey Relational Analysis model based on the Principal Component Analysis is putforward.By building the index system reflecting the system behavior and related factors,the PrincipalComponent Analysis is used to extract the main components,and combined with the variancecontribution rate,the sequence of the main components composite score is computed,then the GreyRelational Analysis method is used to compute the correlational degree of the sequence of maincomponents composite score reflecting the the system behavior and related factors,to achieve thepurpose of the correlation analysis.(2) For the problem of small sample panel data,the Grey Correlation Clustering model for thepanel data is built up.By the analysis to grey clustering coefficient matrix of the differentmoments,and the introduction of Grey Relational analysis idea,the grey clustering coefficient matrixassembled into grey relational coefficient matirx of differet objects belonging to the differentclasses.Based on the principle of maximun correlation coefficient,the object belonging to the somegrey class is determined to achieve the purpose of clustering for the panel data.(3) The Grey Relational Analysis model based on Pricipal Component Analysis and GreyClustering with right weight method based on grey relational analysis are used respectively to solvethe problem of the factor analysis to the financial input-output technology system in Jiangsu and theclustering analysis of the manufacturing-relatied industries in Jiangsu Province.The two empiricalresearch shoe the two models can solve the problems of the analysis to the factor with multi-index andclustering for the panel data associated with small sample,and the effectiveness and feasibility of themodels are aslo validated.