数学建模 based on 清风

层次分析法

原文链接(我的主页):https://rick2pc.github.io/2022/07/30/AHP/
层次分析法,即:“The analytic hierarchy process”, 简称AHP

1.建模比赛中最基础的模型之一;

2.主要解决评价类问题;

3.文章中所有图片来自于清风数学建模课程的课件。

我们用一个例子来介绍这个建模方法:

“填好志愿后,小明同学想出去旅游。在查阅了网上的攻略后,他初步选择 了苏杭、北戴河和桂林三地之一作为目标景点。“

”请你确定评价指标、形成评价体系来为小明同学选择最佳的方案。”

对于解决这种评价类问题,我们需要考虑以下三个问题:

  1. 我们的评价目标是什么?
  2. 我们为了达到这个目标有哪几种可选的方案
  3. 评价的准则或者说指标是什么?(我们根据什么东西来评价好坏)

对于我们的例子,我们的回答:

  1. 为小明同学选择最佳的旅游景点。
  2. 三种,分别是去苏杭、去北戴河和去桂林。
  3. 题目没给相关数据支撑,需要我们查阅相关的资料。

一般而言,前两个问题的答案是显而易见的,第三个问题的答案 需要我们根据题目中的背景材料、常识以及网上搜集到的参考资料进行结合,从中筛选出最合适的指标。

在这里,我们假设5个指标:景点;花费;居住;饮食;交通。

然后我们构建一张权重表:

权重表

要注意, A H P AHP AHP的最终目的其实就是把这张表通过数学的方式(而不是凭直觉)进行填充,而不是小明自己随便填写,“在确定影响某因素的诸因子在该因素中所占的比重时,遇到的主要困难 是这些比重常常不易定量化。此外,当影响某因素的因子较多时,直接 考虑各因子对该因素有多大程度的影响时,常常会因考虑不周全、顾此 失彼而使决策者提出与他实际认为的重要性程度不相一致的数据,甚至 有可能提出一组隐含矛盾的数据。”(司守奎《数学建模算法与应用》)

我们首先来思考一下5个指标之间的权重:

一次性考虑五个之间的关系比较繁杂,我们可以两两进行比较,最终通过两两比较的结果来推算权重。

我们使用1~9表示两两指标之间的重要程度(这里的重要程度也可以理解为满意度),具体数字的含义,见下表:

重要程度

其中,重点要注意一下最后一行那个倒数的含义。

好,我们将根据这个标度,来进行权重的计算:

1. 构建判断矩阵

构造判断矩阵

  1. 首先,对于指标权重,我们会构建一个判断矩阵,再计算权重。如上图,记为 A A A,对应的元素为 a i j a_{ij} aij。该矩阵有如下特点:

    • a i j a_{ij} aij 表示的意思是:与指标 j j j相比, i i i的重要程度。
    • i = j i=j i=j时,两个指标相同,因此同等重要,这就解释了主对角线元素为1。
    • a i j > 0 a_{ij}>0 aij>0且满足 a i j ∗ a j i = 1 a_{ij}*a_{ji}=1 aijaji=1(我们称满足这一条件的矩阵为正互反矩阵)。

    但是,在实际建模的过程中,没有“小明”帮助我们填表,层次分析的这张表是交给“专家”填写的,但在比赛过程中往往很难找找到专家标准,一般只能自己填了,但是在论文中就别说出来了。

  2. 其次,对于三个地点在五个把不同标度分别的得分也是重复上面的操作,例如:

    计算在景色方面苏杭,北戴河,桂林所占的得分,我们也是先构建一个判断矩阵,然后再计算权重:

    准则层判断矩阵

所有判断矩阵

但是,要注意一个问题:

​ 苏杭=A 北戴河=B 桂林=C

​ 苏杭的景色比北戴河好:A>B

​ 苏杭和桂林景色一样好: A = C

​ 北戴河比桂林景色好一点:B > C

这个时候,就会出现问题:如果A = C而A > B,那么C>B与第三条冲突,判断矩阵出现问题

不一致性

判断矩阵出现了矛盾之处,我们称之为不一致现象

何为一致矩阵?简单理解一下就是各行(列)之间成倍数关系,例:

一致性矩阵

若矩阵中每个元素 a i j > 0 a_{ij}> 0 aij>0且满足 a i j ∗ a j i = 1 a_{ij}*a_{ji}=1 aijaji=1,则我们称该矩阵为正互反矩阵。在层次分析法中,我们构造的判断矩阵均是正互反矩阵。

在正互反矩阵满足 a i j ∗ a j k = a i k a_{ij}*a_{jk}=a_{ik} aijajk=aik,则我们称其为一致矩阵

注意:在使用判断矩阵求权重之前,必须对其进行一致性检验。

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-Fc4mYPaJ-1659151534540)(https://raw.githubusercontent.com/Rick2pc/Typora/main/202207301110166.png)]

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-TiUrADbi-1659151534540)(https://raw.githubusercontent.com/Rick2pc/Typora/main/202207301110107.png)]

一致性检验步骤

image-20220730101453645

2. 根据判断矩阵计算权重

方法一:算数平均法求权重
  1. 第一步: 将判断矩阵按照列进行归一化(每一个元素除以其所在列的和);
  2. 第二步: 将归一化的各列相加(按行求和);
  3. 第三步: 将相加得到的向量中每个元素除以 n n n即可得到权重向量。

算数平均法

例:

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-ow2EPXo2-1659151834963)(https://raw.githubusercontent.com/Rick2pc/Typora/main/202207301109075.jpg)]

方法二:几何平均法求权重
  1. 第一步: 将A的元素按照行相乘得到一个新的列向量;
  2. 第二步: 将新的向量的每个分量开n次方;
  3. 第三步:对该列向量进行归一化即可得到权重向量。

几何平均法

方法三:特征值法求权重

假如我们的判断矩阵一致性可以接受,那么我们可以仿照一致矩阵权重的求法。

  1. 第一步:求出矩阵A的最大特征值以及其对应的特征向量;
  2. 第二步:对求出的特征向量进行归一化即可得到我们的权重。

特征值平均法

三种方法说实话感觉没什么区别,写论文的时候选一种就好了,但是最好点明三种方法都算过了,然后在附录里面展现出最后的结果。

3. 计算得分

根据上节的三种方法,我们可以选择一种方法分别计算每个区块的得分,然后填入表中

得分表

每一种颜色就代表了一个判断矩阵

然后根据指标权重和每个地区在5个指标上的权重进行打分,例:

例:计算得分

这样,我们就可以得到三个地区,在五个指标上的得分,进而来选择小明想要去的地方。

4. 总结

层次分析法第一步

1. 层次分析法第二步

2.层次分析法第二步

3.层次分析法第二步

1. 层次分析法第三步

2. 层次分析法第三步

层次分析法第四步

存中…(img-31ciNpwj-1659151534546)]

[外链图片转存中…(img-LVvA6TwZ-1659151534547)]

[外链图片转存中…(img-QVW17jYg-1659151534547)]

[外链图片转存中…(img-vg6YdndN-1659151534547)]

层次分析法的局限性

  • 2
    点赞
  • 13
    收藏
    觉得还不错? 一键收藏
  • 1
    评论
Our planet is getting thirstier and thirstier. Water scarcity has become an increasingly hard but urgent problem. To make contributions to solve the water problems, we proposed a metric model to identify the ability of each country to manage water scarcity, and offered solutions to a country considered water over-loaded. First we developed our metric, Total Scarcity Metric, and divided it into Physical Scarcity Metric (affected by environmental factors and population) and Economic Scarcity Metric (affected by social factors other than population) by the two causes of water scarcity. The detailed factors were selected to make a difference. We made some adjustments to an indicator we found widely-used in the literatures, and determine Physical Scarcity Metric based on it. For Economic Scarcity Metric, we built a factor model with its weight calculated by Grey Relational Analysis. To combine them, we introduced a parameter  revealing the relative emphasis between physical scarcity and economic scarcity of each country. Its value varies by countries, so it’s more proper to discuss it in the country level. Then we used data from 83 countries to verify our model, and found a similar water scarcity distribution compared to UN’s “World Scarcity Map”. By providing sensitivity analysis on  ,we indicated the importance of its selection for each country. Based on that result, we chose Pakistan for further analysis. First we discussed the possible factors accounting for its current water situation, including 2 environmental factors and 5 social factors. We made it clear how and what kind of scarcity they affect. To forecast the water situation in 2030 by our metric, we determined the predicted value of influential factors by Grey Forecasting Model and Regression Analysis with little error. We found Pakistan less susceptible to economic scarcity but more to physical scarcity at that time. Its total water scarcity will be alleviated. Next, we devised an intervention plan to improve the ability of Pakistan to deal with its water scarcity. The plan is made up of physical scarcity plan and economic scarcity plan. Considering the impact of each policy, we analyzed the overall strengths and weaknesses in a larger context. To see how our plan performs, we ran our model again under some hypothesized settings. Although Pakistan performed better with our plan and its economic scarcity can be alleviated, it will still face water scarcity, especially physical water scarcity. In conclusion, Pakistan still has a long way to go.
summary: In this paper, we establish a regression model based on the passing network to evaluate the influence of team structure strategy and opponents’ counter-strategy on the match results. Fortask1,wefirstlistsomeHuskiesmatchstatisticsforthisseasonandanalyzetheteamin brief. Secondly, we construct a passingnetwork based on the number of passes and visualizes the passing network diagram of three games under three different coaches. We use these three diagrams to describe and compare the changes in Huskies’ strategies. After that, we identify network patterns of dyadic and triadic configurations and count 15 kinds of these two configurations in the above three matches, reflecting the structural indicators of the passing network. We also explore time scale and micro scale by giving the change of the team’s centroid over time in the first match and the Huskies’ 4 positions heat map over the season. For task 2, we construct the regression model not only introducing the basic data representing Huskies’ and opponents’ ability, but also extracting six independent variables from the indicators of the passing network into the model. Considering opponents’ counter-strategies, we also introduce the product interaction term between opponents’ data and network structure indicators. Through the training of regression model, we can judge whether the independent variables introduced have influence, what kind of influence and how much influence the independent variables introduced have on the result of the match. For task 3, by bringing in data for training, the model leaves 10 variables including interactionterms. Inordertoverifytheaccuracyofthemodel,weuseLeaveOneOutcrossvalidation, andthepredictedaccuracyoftheraceresultreached71.05%. Then,basedonthetrainedmodel, we point out the effective structural strategies Huskies currently have, such as the strong connection between the core players. Meanwhile, we also give specific advice for Huskies team to improve team success, such as the emphasis on triadic configurations among players. Fortask4,weextendthemodelappliedtohuskiestoallteamworkscenariosandintroduce the IPOI model. The IPOI model conducts multi-level induction of influencing factors and selection of assessment indicators from the four aspects of team input, process, output and reinput, taking into account team construction, operation, management, feedback and other aspects. WeconsiderthattheexistingHuskiemodelisprogresspartofIPOImodel,andweadd the evaluation system of input, output and reinput part, taking the university scientific research team modeling as an example. Insummary,ourmodelispracticalandreliableforhandlingnetwork-basedteamworkproblems in society. Keywords: football strategy, network science, regression analysis, IPOI model.

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值