The Imputation Methods For Missing Values In Compositional Data Based On Kernel Function

Due to the awareness of scientific progress,analysis and study of scientific spirit gradually thorough popular feeling,modern life,often need to face the data collecting and processing in order to finish daily work more efficiently.In all possible data,data is a kind of meet the complex features of multidimensional data,commonly used to study in a whole under the specified factors proportion relationship between the parts.Along with the economic development level enhances unceasingly,from all walks of life more and more aware of the benefits of accurate data statistics,composition data so also more and more widely applied.Practical problems,however,we found that collect statistical data often has deficiencies,such as the invalid or blank information in the questionnaire,collecting the omission,etc all make a missing data.Statistical quality will be affected by the missing data,which leads to the estimation deviation,produce adverse results.So we hope we can comp-lete data,so it is particularly important to the completion of missing data.In terms of missing data processing at home and abroad,a lot of achievements in this paper,on the basis of predecessors' research,try to fill the missing value,the method of using the kernel function study compared the advantages and disadvantages of the methods.This paper is divided into five chapters:the first chapter illustrates the research signify-cance of this paper,this paper expounds the current research background,research status at home and abroad,and made the outline of some of the basic situation.The second chapter briefly describes the basic concept of component data,as well as the need to use the relevant knowledge,large operation in the process of research is described,and some of the existing method is introduced.The third chapter is this paper,several kinds of component is proposed based on kernel function data missing value fill method,illustrates the reasons,process and the concrete implementation steps of this method.Chapter iv based on the proposed several missing value fill method based on kernel function is compared with existing common method of simulation experiment,the experimental results,and to analyze a real data,to verify that the method is feasible.Last chapter summarizes the research conclusion of this paper and the prospect of future research.