数学建模 线性规划

一.概念
1.数学规划:

人们经常会遇到如何利用现有资源进行生产,以取得最大经济效益的问题.这类问题形成了运筹学的1个重要分支——数学规划,即研究如何在一定的资源和条件限制下,将某个目标最大或最小化

2.线性规划
(1)概念:

线性规划(Linear Programming;LP)是数学规划的1个重要分支,专门研究线性规划问题,即在1组线性约束条件的限制下,求1个线性目标函数的最大或最小值的问题

(2)一般形式:

一般线性规划问题的(数学)标准型为:$$\begin{gathered} max/minz=\sum _{j=1}^n{c}_j{x}_j(1.1) \\ s.t.\{\begin{array}{l}{b}_{i1}\le \sum _{j=1}^n{a}_{ij}{x}_j\le b_{i2}(i=1,2...m)\\ l\le x_j\le u(j=1,2...n)\end{array}(1.2) \end{gathered}$$其中式(1.1)称为问题的目标函数;式(1.2)称为问题的约束条件,记为$s.t.$(即subject to);称满足式(1.2)的解$x=[x_1,x_2...x_n{]}^T$为线性规划问题的可行解,所有可行解构成的集合称为问题的可行域,记为$R$;而使式(1.1)达到最大或最小值的可行解称为最优解

二.利用MATLAB求解
1.标准形式:

MATLAB规定线性规划问题的标准形式为:$$\begin{gathered} \underset{x}{\min }{f}^{T}x \\ s.t.\{\begin{array}{l}A\cdot x\le b\\ Aeq\cdot x=beq\\ lb\le x\le ub\end{array} \end{gathered}$$其中$f,x,b,beq,lb,ub$为列向量.$f$称为价值向量,$b$称为资源向量;$A,Aeq$为矩阵

2.求解:

MATLAB中使用linprog()求解线性规划问题(详情参见 MATLAB.函数 部分)

例如:$$\begin{gathered} maxz=2x_1+3x_2-5x_3 \\ s.t.\{\begin{array}{l}{x}_1+x_2+x_3=7\\ 2x_1-5x_2+x_3\ge 10\\ x_1+3x_2+x_3\le 12\\ x_1,x_2,x_3\le 0\end{array} \end{gathered}$$的标准形式为:$$\begin{gathered} minw=-2x_1-3x_2+5x_3 \\ s.t.\{\begin{array}{l}\left[\begin{array}{ccc}-2& 5& -1\\ 1& 3& 1\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]\le \left[\begin{array}{c}-10\\ 12\end{array}\right]\\ \left[\begin{array}{ccc}1& 1& 1\end{array}\right]\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]=7\\ \left[\begin{array}{c}{x}_1\\ x_2\\ x_3\end{array}\right]\ge \left[\begin{array}{c}0\\ 0\\ 0\end{array}\right]\end{array} \end{gathered}$$求解代码为:

>> f=[-2;-3;5];
>> a=[-2,5,-1;1,3,1];
>> b=[-10;12];
>> aeq=[1,1,1];
>> beq=7;
>> [x,y]=linprog(f,a,b,aeq,beq,zeros(3,1));

Optimal solution found.

>> x,-y

x =

    6.4286
    0.5714
         0


ans =

   14.5714

三.可转换为线性规划的问题
1.绝对值求和类:

现有数学规划问题:$$\begin{gathered} min\sum _{i=1}^n{a}_i|x_i| \\ s.t.Ax\le b \end{gathered}$$其中$x=[x_1,x_2...x_n{]}^T$;$A,b$为相应维数的矩阵和向量.现在取$$u_i=\frac{x_i+|x_i|}{2},v_i=\frac{|x_i|-x_i}{2}(i=1,2...n)$$并记$u=[u_1,u_2...u_n{]}^T,v=[v_1,v_2...v_n{]}^T$,从而把上述问题转换为$$\begin{gathered} min\sum _{i=1}^n{a}_i(u_i+v_i) \\ s.t.\{\begin{array}{l}A(u-v)\le b\\ u,v\ge 0\end{array} \end{gathered}$$再进一步改写为$$\begin{gathered} min\sum _{i=1}^n{a}_i(u_i+v_i) \\ s.t.\{\begin{array}{l}\left[\begin{array}{cc}A& -A\end{array}\right]\left[\begin{array}{c}u\\ v\end{array}\right]\le b\\ u,v\ge 0\end{array} \end{gathered}$$

求解下述数学规划问题:$$\begin{gathered} minz=|x_1|+2|x_2|+3|x_3|+4|x_4| \\ s.t.\{\begin{array}{l}{x}_1-x_2-x_3+x_4\le -2\\ x_1-x_2+x_3-3x_4\le -1\\ x_1-x_2-2x_3+3x_4\le -\frac{1}{2}\end{array} \end{gathered}$$做变量替换$u_i=\frac{x_i+|x_i|}{2},v_i=\frac{|x_i|-x_i}{2}(i=1,2,3,4)$,并把新变量重新排序成1维向量$y={\left[\begin{array}{cc}u& v\end{array}\right]}^T=[u_1...u_4,v_1...v_4{]}^T$,则$$\begin{gathered} minz=c^{T}y \\ s.t.\{\begin{array}{l}\left[\begin{array}{cc}A& -A\end{array}\right]\left[\begin{array}{c}u\\ v\end{array}\right]\le b\\ y\ge 0\end{array} \end{gathered}$$其中$$\begin{gathered} c={\left[\begin{array}{cccccccc}1& 2& 3& 4& 1& 2& 3& 4\end{array}\right]}^T \\ b={\left[\begin{array}{ccc}-2& -1& -\frac{1}{2}\end{array}\right]}^T \\ A=\left[\begin{array}{cccc}1& -1& -1& 1\\ 1& -1& 1& -3\\ 1& -1& -2& 3\end{array}\right] \end{gathered}$$求解代码为:

>> c=1:4;
>> c=[c,c]';
>> a=[1,-1,-1,1;1,-1,1,-3;1,-1,-2,3];
>> a=[a,-a];
>> b=[-2,-1,-1/2]';
>> [y,z]=linprog(c,a,b,[],[],zeros(8,1))

Optimal solution found.


y =

     0
     0
     0
     0
     2
     0
     0
     0


z =

    2.0000

>> x=y(1:4)-y(5:end)

x =

    -2
     0
     0
     0

2.多重最值类:

现有数学规划问题: m i n x i { m a x y i ∣   ε i ∣ } \mathop{\mathbb{min}}\limits_{x_i}\{\mathop{\mathbb{max}}\limits_{y_i}|\,ε_i|\} xi​min​{yi​max​∣εi​∣}其中${\epsilon }_i=x_i-y_i$.取$v=\underset{y_i}{\max }|{\epsilon }_i|$,上述问题就转换为:$$\begin{gathered} minv \\ s.t.\{\begin{array}{l}{x}_i-y_i\le v\\ y_i-x_i\le v\end{array} \end{gathered}$$

四.投资的风险和收益

假设市场上有$n$种资产可供选择,分别记为$s_i(i=1,2...n)$.现用数额为$M$的相当大的资金进行投资,这$n$种资产的平均收益率分别为$r_i$,风险损失率为$q_i$.投资越分散,总风险就越小,总风险可用投资的$s_i$中最大的1个风险来衡量.购买$s_i$时交易费的费率为$p_i$,但当购买额不超过$u_i$时,交易费均按购买$u_i$计算.另外,同期银行存款(记为$s_0$)利率为$r_0$,没有交易费和风险(即$p_0=q_0=0$).设投资在$s_i$上的资金额为$x_i$,各种资产间相互独立.要求设计1种投资组合方案,使净收益尽可能大,总风险尽可能小

总风险 A = m a x { q i x i   ∣   i = 1 , 2... n } A=max\{q_ix_i\,|\,i=1,2...n\} A=max{qi​xi​∣i=1,2...n}.用于$s_i$的交易费$P_i=\{\begin{array}{l}{p}_i{x}_i(x_i>u_i)\\ p_i{u}_i(x_i\le u_i)\end{array}$.由于$p_i{u}_i<u_i<<M$,$s_i$的净收益$R_i\approx (r_i-p_i)x_i$
从而这是1个多目标规划模型 { m a x ∑ i = 0 n ( r i − p i ) x i m i n { m a x 1 ≤ i ≤ n { q i x i } } s . t . { ∑ i = 0 n ( 1 + p i ) x i = M x i ≥ 0     \begin{cases}max\displaystyle\sum_{i=0}^n(r_i-p_i)x_i\\min\{\mathop{\mathbb{max}}\limits_{1≤i≤n}\{q_ix_i\}\}\end{cases}\\s.t.\begin{cases}\displaystyle\sum_{i=0}^n(1+p_i)x_i=M\\x_i≥0\end{cases}\quad\:\:\: ⎩⎪⎪⎨⎪⎪⎧​maxi=0∑n​(ri​−pi​)xi​min{1≤i≤nmax​{qi​xi​}}​s.t.⎩⎪⎨⎪⎧​i=0∑n​(1+pi​)xi​=Mxi​≥0​

①若希望在单项投资的风险率$($即$\frac{q_i{x}_i}{M})$不超过$a$的情况下获得最大收益,则模型简化为$$\begin{gathered} max\sum _{i=0}^n(r_i-p_i)x_i \\ s.t.\{\begin{array}{l}\frac{q_i{x}_i}{M}\le a\\ \sum _{i=0}^n(1+p_i)x_i=M\\ x_i\ge 0\end{array} \end{gathered}$$②若希望在总盈利不少于$k$的情况下使风险最小,则模型简化为 m i n { m a x 1 ≤ i ≤ n { q i x i } } s . t . { ∑ i = 0 n ( r i − p i ) x i ≥ k ∑ i = 0 n ( 1 + p i ) x i = M x i ≥ 0 min\{\mathop{\mathbb{max}}\limits_{1≤i≤n}\{q_ix_i\}\}\\s.t.\begin{cases}\displaystyle\sum_{i=0}^n(r_i-p_i)x_i≥k\\\displaystyle\sum_{i=0}^n(1+p_i)x_i=M\\x_i≥0\end{cases} min{1≤i≤nmax​{qi​xi​}}s.t.⎩⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎧​i=0∑n​(ri​−pi​)xi​≥ki=0∑n​(1+pi​)xi​=Mxi​≥0​③若希望权衡风险和预期收益这2方面,且对风险,收益分别赋权为$0<s\le 1,1-s(s$称为投资偏好系数$)$,则模型简化为 m i n    s { m a x 1 ≤ i ≤ n { q i x i } } − ( 1 − s ) ∑ i = 0 n ( r i − p i ) x i s . t . { ∑ i = 0 n ( 1 + p i ) x i = M x i ≥ 0 min\,\:s\{\mathop{\mathbb{max}}\limits_{1≤i≤n}\{q_ix_i\}\}-(1-s)\displaystyle\sum_{i=0}^n(r_i-p_i)x_i\\s.t.\begin{cases}\displaystyle\sum_{i=0}^n(1+p_i)x_i=M\\x_i≥0\end{cases} mins{1≤i≤nmax​{qi​xi​}}−(1−s)i=0∑n​(ri​−pi​)xi​s.t.⎩⎪⎨⎪⎧​i=0∑n​(1+pi​)xi​=Mxi​≥0​