您好,欢迎访问三七文档
一、最大特征值和特征向量的近似计算(方根法)计算的主要步骤:1、计算判断矩阵A的每一行元素乘积1,1,2,,.niijjMain2.计算Mi的n次方根.niiWM3.若iW标准化为1,iinjj则iW为所求特征向量。4.计算最大特征值max1().niiiAWnW式中AWi表示向量AW的第i个分量。例如1110.711/51/3535135131531/3113113每行之乘积0.4120.1052.4660.637,10.258iM球的三次方根标准化即权系数为(0.105,0.637,0.258)TW0.10511/51/30.1050.3180.637.5130.6371.9360.25831/310.2580.785WAW再求最大特征值的近似值3max1()0.3181.9360.7853.03730.10530.63730.258iiiAWnW二、迭代法建立n阶方阵.()ijnnAa按下列方法求向量迭代序列:01.111nnenn'ke为1kAe的n个分量之和可以证明,迭代的维列向量序列{}ke记其极限为e,且记收敛。121.nne则权系数可取,1,2,,iiin在具体计算中,当ek与ek-1接近到一定程度时,我们就取e=ek例如0111/51/3111/3,1/35311/3Ae101111/51/30.733'111/31/30.778,'0.7330.77834.5115311/33eAee1110.7330.162'10.7780.172'4.51130.665eee212111/50.1620.467'111/30.1720.565,'0.4670.5651.9913.0145310.6651.991eAee2220.4670.155'10.5650.184'3.0141.9910.661eee323111/50.1550.471'111/30.1840.559,'0.4710.5591.9883.0185310.6611.988eAee3330.4710.156'10.5590.185'3.0181.9880.659eee434111/50.1560.473'111/30.1850.561,'0.4730.5611.9943.0285310.6591.994eAee4440.4730.156'10.5610.185'3.0281.9940.659eee由于e4=e3,迭代经过4次中止,权系数是1230.156,0.185,0.659.相应的综合评价公式是1230.1560.1850.659yxxx三、和法1、计算判断矩阵A的每一列归一化1/nijijijiaa2.按行求和1niijjw3.归一化1/niiii12(,,,)Tn则Wi为所求特征向量。4.计算AW5、计算最大特征值得近似值max1().niiiAWnW式中(AW)i表示向量AW的第i个分量。例某厂准备购买一台计算机,希望功能强,价格低,维护容易。现有A、B、C三种机型可供选择。其中A的性能较好,价格一般,维护一般水平;B的性能最好,价格较贵,维护也只需一般水平;C的性能差,但价格便宜,容易维护。试用层次分析法进行决策分析。解:1、明确问题;2、建立层次结构;先构造层次结构,如下图所示。购买一台满意的计算机G功能强S1价格低S2CAB易维护S3目标层方案层准则层对于三个准则(S1,S2,S3)关于目标G的优先顺序,根据讨论,该厂在计算机应用上首先要求功能强,其次要求易维护,再次才是价格低。其判断矩阵如下表所示。3、构造判断矩阵表一GS1S2S3S1S2S311/51/351331/314、层次单排序及其一致性检验(用方根法计算这三个准则关于目标的排序权值)12315,0.667,1MMM333123152.446,0.6670.405,11标准化:12.4462.4460.6372.4460.40513.871W310.258.2.4460.4051W20.4050.1052.4460.4051W则iW为所求特征向量。计算最大特征值max1().niiiAWnW式中()iAW表示向量AW的第i个分量。一致检验结果为0.6371530.6370.105.1/511/30.1050.2581/3310.258WAW10.637+50.105+30.2581.9361/50.637+10.105+1/30.2580.3181/30.637+30.105+10.2580.7853max1()1.9360.3180.78530.63730.10530.258iiiAWnW1.9360.3180.7853.03851.9110.3150.774判断矩阵的一致性指标CI为max3.038530.0192,131nCIn判断矩阵的同阶平均随机一致性指标RI为(3)0.58RI判断矩阵的一致性指标CI与判断矩阵的同阶平均随机一致性指标RI之比称为随机一致性比率,记为CR,当0.01920.03320.100.58CICRRI时,即认为判断矩阵具有满意的一致性,否则就需要调整判断矩阵,使其具有满意的一致性。5、层次总排序这一过程是最高层次到最低层次逐层进行的。若上一层次A包含m个元素12,,,,mAAA其层次总排序权值分别为12,,,,maaa次B包含n个元素下一层12,,,,nBBB它们对于元素jA的层次单排序权值分别为12,,,jjnjbbb(当kB与jA无关系时,0ijb)排序权值如表二给出。。此时B层次总层次A1A2……AmB层次总排序权重a1a2……amB1B2┋Bnb11b12……b1mw1b21b22……b2mw2┋┋┋┋bn1bn2……bnmwn注:1(1,2,,)mijijjwabin同样,三个方案对于各个准则的判断矩阵以及运算所得的结果见三、四、五、六。表三1SABCWABC11/424181/21/810.18180.72720.09100.181811/420.18180.7272.4180.72720.09101/21/810.0910WAW10.1818+0.72721/4+20.09100.545640.1818+0.72721+80.09102.18241/20.1818+0.72721/8+10.09100.27283max1()0.54562.18240.272830.181830.727230.091iiiAWnW0.54562.18240.27283.00.54542.18160.273对准则1S(功能强)来说:3.030.0,131nCInmax3.0max00.00.10.0.58CICRRI即认为判断矩阵具有满意的一致性。表四2SABCWABC141/31/411/83810.25590.07330.6708对准则2S(价格低)来说:3.018,0.0091,0.01570.10.CICRmax即认为判断矩阵具有满意的一致性。表五3SABCWABC111/3111/53510.18510.15620.6587对准则3S(价格低)来说:3.029,0.0145,0.0250.10.CICRmax即认为判断矩阵具有满意的一致性。表六总排序权值0.6370.1050.258ABC0.18180.25590.18510.72720.07330.15620.09100.67080.65870.18180.72720.09101S2S3SG0.18180.6370.25590.1050.18510.2580.18180.72720.6370.07330.1050.15620.2580.51120.0910.6370.67080.1050.65870.2580.29840.63700.1050.00910.2580.01450.004710.004710.58,0.00810.10.0.58CICIRICRRI即认为判断矩阵具有满意的一致性。从以上结果可知,B型计算机从综合评价来看是最满意的备选机型。
本文标题:层次分析法的计算
链接地址:https://www.777doc.com/doc-2055280 .html