基于多元定量资料研究样品间关系的综合评价方法的比较研究

发布时间：2019-04-02 14:58

【摘要】：目的本研究拟解决目前国内外综合评价研究中待解决的两个重要问题:(1)提出一种更具有合理性的综合评价方法;(2)回答多种综合评价方法评价结果中何者更合理的问题。方法通过广泛查阅文献,参考国内外相关研究成果,将全部可能用于综合评价的数据结果进行总结和归纳。通过对教材和文章的学习,明确阐释样品聚类分析与传统综合评价方法的区别与联系,并对传统综合评价方法和多元统计分析方法进行全面的梳理。通过借鉴投影寻踪聚类分析中投影指标函数的构造原理以及任何两个样品间易于确定先后顺序的基本常识,提出新方法和判定综合评价方法优劣的判定标准。选择已采用投影寻踪聚类分析处理的五篇文献案例中的原始数据,对其再进行八种其他综合评价方法的评价效果验证。基于对两种较复杂且可能有实用价值的综合评价方法(即模糊综合评价法和灰色关联度分析法)的原理的学习,编写补充实现前述提及的两种综合评价方法计算的SAS程序。结果总结出可能出现的四种数据结构;新构建出一种新的综合评价分析方法,即“基于标准化变换的求和法”;新提出判定综合评价方法优劣的两条评判标准;基于投影寻踪聚类分析的五个数据集进行八种综合评价方法的评价效果验证。结论(1)首次建立了一种新的综合评价方法,即基于标准化变换的求和法。该方法通用于Ⅰ型数据结构,灵活方便。(2)提出了判定不同综合评价方法优劣的标准。借鉴投影指标函数的构造原理以及易于直观判断任何两个样品排列顺序的事实。解决了目前文献中没有明确的判断综合评价方法优劣的难题。(3)在目前实现综合评价方法计算的软件不够齐全的情况下,编写了模糊综合评价法和灰色关联度分析法的SAS程序。(4)研究发现,传统综合评价方法(包括本文的基于标准差变换的求和法)的结果合理性普遍优于间接实现综合评价的几个多元统计分析方法,其中,又以本文方法为最合理。
[Abstract]:Objective to solve two important problems in the research of comprehensive evaluation at home and abroad: (1) to propose a more reasonable comprehensive evaluation method; (2) to answer the question which is more reasonable in the evaluation results of various comprehensive evaluation methods. Methods all the data which may be used for comprehensive evaluation were summarized and summarized by referring to the related research results at home and abroad through extensive reference to the literature and the related research results at home and abroad. Through the study of teaching materials and articles, the differences and relations between sample clustering analysis and traditional comprehensive evaluation methods are clearly explained, and the traditional comprehensive evaluation methods and multivariate statistical analysis methods are comprehensively combed out. Based on the construction principle of projection index function in projection pursuit cluster analysis and the basic knowledge that it is easy to determine the sequence between any two samples, a new method and a criterion for judging the merits and demerits of comprehensive evaluation methods are proposed. The original data from five literature cases which have been processed by projection pursuit cluster analysis are selected, and the evaluation results of eight other comprehensive evaluation methods are verified. Based on the study of the principles of two comprehensive evaluation methods (fuzzy comprehensive evaluation method and grey relational degree analysis method) which are complex and may be of practical value, the SAS program which supplements the calculation of the two comprehensive evaluation methods mentioned above is compiled. Results four possible data structures were summarized, a new comprehensive evaluation and analysis method, the summation method based on standardized transformation, two new criteria for judging the merits and demerits of comprehensive evaluation methods, and two new evaluation criteria were proposed, and a new comprehensive evaluation method, called "sum-up method based on standardized transformation", was proposed. Five data sets based on projection pursuit clustering analysis are used to verify the evaluation results of eight comprehensive evaluation methods. Conclusion (1) A new comprehensive evaluation method, the sum method based on standardized transformation, is established for the first time. The method is universal to type I data structure and is flexible and convenient. (2) the criteria for judging the merits and demerits of different synthetic evaluation methods are proposed. Reference is made to the construction principle of projection index function and the fact that it is easy to judge the order of any two samples directly. The problem of judging the advantages and disadvantages of the comprehensive evaluation method in the literature is solved. (3) under the condition that the software for the calculation of the comprehensive evaluation method is not complete at present, The SAS program of fuzzy comprehensive evaluation method and grey relational degree analysis method is compiled. (4) it is found that The rationality of the traditional comprehensive evaluation method (including the sum method based on the standard deviation transformation) is generally better than that of several multivariate statistical analysis methods to realize the comprehensive evaluation indirectly. Among them, the method in this paper is the most reasonable.
【学位授予单位】：中国人民解放军军事医学科学院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O212

【参考文献】