安玲学记（238）——精读硕士论文4.3指标权重的确定（2）

摘要：In this issue, the tweet will introduce the 4.3 determination of evaluation index weights (2) of the master's paper "Research on e

分享兴趣，传播快乐，

增长见闻，留下美好。

亲爱的您，这里是LearingYard学苑！

今天小编为大家带来“精读硕士论文《A公司数据平台项目软件开发服务供应商评价研究》的4.3评价指标权重的确定（2）”。

欢迎您的访问！

Share interest, spread happiness,

increase knowledge, and leave beautiful.

Dear, this is the LearingYard Academy!

Today, the editor brings the "4.3 determination of evaluation index weights (2) of software projects of the master's paper 'Research on evaluation of software development service providers of A company's data platform proiect'".

Welcome to visit!

一、内容摘要（Content summary）

本期推文将从思维导图、精读内容、知识补充三个方面介绍精读硕士论文《A公司数据平台项目软件开发服务供应商评价研究》的4.3评价指标权重的确定（2）。

In this issue, the tweet will introduce the 4.3 determination of evaluation index weights (2) of the master's paper "Research on evaluation of software development service providers of A company's data platform proiect" from three aspects: mind map, detailed reading content, and additional knowledge supplementation.

二、思维导图（Mind Mapping）

三、精读内容（Detailed Reading Content）

该部分聚焦数据平台项目软件开发服务供应商评价模型的构建，分为指标权重确定和权重评价应用。上期推文分享了权重确定步骤的前三步，即：构建层次结构模型、权威评价专家的选择、构造判断（成对比较）矩阵。本次推文将分享指标权重确定的剩余部分内容。

This part focuses on the construction of evaluation model of software development service provider of data platform project, which is divided into index weight determination and weight evaluation application. The previous tweet shared the first three steps of the weight determination step, namely: building a hierarchical structure model, selecting an authoritative evaluation expert, and constructing a judgment (pair comparison) matrix. This tweet will share the rest of the metric weight determination.

第四步为指标权重计算与一致性检验。以一位专家数据为例，将一级指标的5个指标要素构成两两比较矩阵，运用1-9标度法得出判断矩阵，然后对判断矩阵的列向量归一化处理后行求和再归一化，得到近似特征向量即权重向量，再得出最大特征根，进而算出一致性指标，依据五阶矩阵随机一致性指标算出随机一致性比率，最后判定该专家的判断数据符合一致性要求。

The fourth step is the index weight calculation and consistency test. Taking the data of an expert as an example, the five index elements of the first-level index are formed into a ptwo comparison matrix, and the judgment matrix is obtained by using the 1-9 scale method. After the normalization of the column vectors of the judgment matrix, the sum of the rows is normalized to obtain the approximate feature vector, that is, the weight vector, and the maximum feature root is obtained to calculate the consistency index. According to the random consistency index of the fifth-order matrix, the random consistency ratio is calculated, and the expert's judgment data is judged to meet the consistency requirements.

第五步为利用软件一致性检验结果验证。为避免人为计算错误，作者使用yaahp软件对前文同一专家的判断矩阵进行一致性检验以保证数据可对比性。结果表明，人为计算和软件计算出的一致性检验结果一致，在判定矩阵一致性检验精度要求不高时，两种方法的计算结果和矩阵一致性检验结果相同，均可用于判断矩阵一致性情况。

The fifth step is to use the software consistency test results to verify. In order to avoid human calculation errors, the author uses yaahp software to carry out consistency test on the judgment matrix of the same expert mentioned above to ensure data comparability. The results show that the consistency test results obtained by human calculation and software are the same. When the accuracy requirement of matrix consistency test is not high, the calculation results of the two methods are the same as that of matrix consistency test, and both methods can be used to judge matrix consistency.

第六步为利用软件分析得出权重。经软件自动修正、补全和计算，得出评价模型各元素权重数据。一级指标对评价目标的权重排序为企业开发能力＞类似项目经验＞企业资质＞对项目的理解＞费用；二级指标中，开发人员实力、建设案例及替换（割接、重构）案例被专家认为是最重要的三项评价指标。最后，经计算整理得到二级指标权重分配，各二级指标权重加总后得到一级指标权重（准则层权重）。

The sixth step is to use software analysis to get the weight. The weight data of each element of the evaluation model are obtained by software automatic correction, completion and calculation. The weights of the first level index on the evaluation objectives are as follows: enterprise development ability > similar project experience > enterprise qualification > understanding of the project > cost; Among the secondary indicators, the strength of developers, construction cases and replacement (cut, reconfiguration) cases are considered by experts to be the most important three evaluation indicators. Finally, the weight distribution of the secondary index is obtained after calculation, and the weight of the primary index is obtained after the weight of each secondary index is added together (the weight of the criterion layer).

四、知识补充——一致性检验（ Knowledge Supplement - Consistency Test）

一致性检验的目的在于比较不同方法得到的结果是否具有一致性。检验一致性的方法有很多比如：Kappa检验、ICC组内相关系数、Kendall W协调系数等。每种方法的功能侧重，数据要求都略有不同。

The purpose of consistency test is to compare whether the results obtained by different methods are consistent. There are many methods to test consistency, such as Kappa test, ICC intra-group correlation coefficient, Kendall W coordination coefficient, etc. Each approach has slightly different functional priorities and data requirements.