Non-negative Matrix Factorization And Its Application In The Proficiency Testing

Text data is typically processing in the form of a matrix by computer, and the feature data matrix with high dimensional sparse, while non-negative matrix factorization method is a new solution to reduce the dimension of the original data. The method is based on the traditional robust statistical analysis to validation data processing testing with so many problems in the past. Such as, it provides less information, it can’t describe correlation of the laboratory data, only to analyze a single project when test object is multivariate, lack of overall judgment and description, and it can’t adapt the content and objects increasing trend. To solve such problems, the paper will be introduced non-negative matrix factorization (NMF) to the proficiency testing data processing. With five practical examples as example to deal with their data, and the results of the analysis are compared with the robust statistics Z score method. The research indicates that this method can analyze data with multi-level and multi-angle. The laboratory will be classified by means of test results and the similarity, it is not only easy to find and analyze the problem of the detection process, but also conducive to the understanding and evaluation of testing laboratories of their own test level.The dissertation is divided into four chapters.Chapter One is an introduction, which contains background, significance of the research, Proficiency Testing and main works.Chapter two presents the study of robust statistics Z score method. Briefly introduces the robust statistics Z score of the principle and the application in Proficiency Testing, it is currently judging lab method for China’s proficiency testing, which has carried on the example analysis. At the same time, it is pointed out the lack of robust statistics Z score method.The third chapter is the study of non-negative matrix factorization, it mainly introduces the non-negative matrix factorization of the definition, principle and basic algorithm, this method is reasonable and effective of Proficiency Testing with five practical examples, when the result is confliction between the Z-scores determine of the determination of the results with the corresponding testing standards, Non-negative Matrix Factorization is the choice of statistical methods to prevent misjudgment of the laboratory test results as complementary tools for inspection. At the same time, NMF is able to detect macro analysis and microanalysis testing. Robust statistical methods combined with Non-negative Matrix Factorization can provide a basis of the critical value near the laboratory judgment and attribution.Chapter four is the conclusion and outlook. the content of the paper is a summary of the application with non-negative matrix factorization in proficiency testing, and it puts forward the research contents of later.