Study On The Grey Panel Data Cluster Method And Its Application

Nowadays,with the risky elements in the process of social development and fierce competition of the market,both the economic system and the social system often generate lots of data during the working period.There are full of uncertain factors in the operation system,such as incompleteness information,preference information,inaccuracy information,noise data,ambiguity information,grey information,etc..For optimizing the system and improving the operational efficiency,It's a good choice for the whole system to analyze and deal with these uncertain situations by data mining effectively using scientific methods.In the real economy and society,due to the coexistence of risks and uncertainties,the massive data has the feature of high dimensionality and significant correlations.If we study the problems using traditional analysis methods,it will not meet the feature of data and existing research conditions,it will further influence deeply on the data mining in the process of information identification.As a result,we cannot obtain the effective high-quality deep knowledge to help scientific decisionmaking.Panel data own the both features of the cross-sectional data and time series data,which can be used to describe the systematic and dynamics of the sample in real conditions.Therefore,this paper aims to solve the inaccurate clustering results and lack of information between objects in the process of clustering evaluation of existing panel data.To effectively mine the difference information between the clustered objects' relational information and the spatiotemporal feature attributes,based on the multiple information features of panel data,we introduce the grey target theory to develop a new multi-index adaptive weight grey panel data clustering method based on the spatial-temporal feature attributes of panel data.The evaluation methods are used to improve the information mining efficiency between clustering objects and obtain more accurate clustering results,thus generating high-quality deep knowledge to help decision-maker in decision-making.The proposed new grey panel data clustering method avoids the selection of initial points,the determination of the number of clustering categories,and the constraint of attribute weights.It can not only effectively classify the temporal and spatial feature attribute categories of panel data containing grey information,but also degenerate into traditional twodimensional traditional panel data clustering methods,thereby improving the information mining efficiency between clustering objects,making the final clustering results more accurate.To illustrate the effectiveness and rationality of the proposed new method,we design a test with run-time,cluster accuracy and error rate compared to the traditional clustering method,such as fuzzy method and K-means method.It's very important for the manifestation of collaborative innovation and comprehensive development of universities and enterprises.However,due to the limitations of human cognition and the uncertainty of the development of things,the phenomenon of “two skins” in theory and practice existed in the development of cooperation between universities and universities in China,which was highlighted by the low level of cooperation in production,education,research,and lack of evaluation process.Based on the scientific basis,it is impossible to provide targeted countermeasures and suggestions on the characteristics and specific problems of universities in different provinces.In a nutshell,this paper adopts the proposed new grey panel data clustering method and on the premise of collecting relevant data and analyzing the background of the cooperation between universities and companies in provinces in China,it analyzes the cooperation of industry,university and research institutes in China.We find the research results as following:1)Considering the clustering result aboard of the panel data and information mining completely in the system,we propose a new multi-index adaptive weight grey panel data clustering method to research the cluster problem describing the trending of the objects in future;2)We test the run-time,cluster accuracy and error rate of the new method we proposed compared to the traditional clustering method,such as fuzzy method and K-means method.The test results shows the advance of the new method we developed;3)Based on the data which are sourced from the National university technology Statistical compilation,we analyze the different province's outcome in the absolute level,incremental level and volatility level using the multi-index adaptive weight grey panel data clustering method.The method solving this case help us avoid setting thresholds according to the size of measurement values to determine the division of attribute categories in each province.Also,the clustering results obtained through spatialtemporal feature attribute information mining provide more detailed and accurate clustering results,so as to give suggestions for different provinces.Moreover,more specific and effective promotion strategies could be given based on the actual situation in the province,it can effectively reduce the risk of increased risk and waste of resources due to policy mistakes.In the end,we put forward targeted countermeasures for the cooperation of companies,universities and research institutes.