今天看到一篇文章介绍如何用excel建模对ROI 进行规划求解。
蓝鲸的网站分析笔记
- 成本 Cost
- 每次点击费用 CPC
- 点击量 \[clickRate = \frac{cost}{CPC}\]
- 转化率 conversionRatio
- 购买量 \[Purchaseamount = \frac{Cost*conversionRatio}{CPC}\]
- 客单价 perCustomerTransaction
- 利润率 rateofProfit
- 利润 \[profit = Purchaseamount * perCustomerTransaction * rateofProfit\]
- 投资回报率 \[ ROI = \frac{profit}{Cost}\]
涉及 多元非线性规划分析求解 ,R 中提供了Rdonlp2()包。
library(Rdonlp2)
p<-c(7,0.01,1000) # 迭代初始值
par.l=c(7,0,0);par.u=c(Inf,0.04,Inf) #自变量定义域约束
fn=function(x){
0.18*x[2]*x[3]/x[1]
} #目标函数
nlcon=function(x){
0.18*x[2]*x[3]/x[1]
}
nlin.l=1.2;nlin.u=1.2 #构成非线性约束
ret<-donlp2(p,fn,par.u=par.u, par.l=par.l,
nlin=list(nlcon), nlin.u=nlin.u, nlin.l=nlin.l)
ret
运行后的结果如下
ret
$par #DONLP2 的返回值 和Excel求解值相同
[1] 7.000 0.040 1166.667
$gradf #梯度
[1] -0.171428571 30.000000000 0.001028571
$u # vector of lagrange multipliers for constraints 约束拉格朗日乘子的矢量
[1] 5.551115e-17 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
[7] 1.000000e+00 0.000000e+00
$w #vector of penalty term 惩罚因子向量
[1] 1.0 1.0 1.0 1.0 1.0 1.0 2.2 1.0
$step.nr # total number of iterations 迭代次数
[1] 2
$fx # the value of objective function fn 目标函数fn的结果
[1] 1.2
$scf # scaling of fn 尺度
[1] 1
$psi # psi the weighted penalty term
[1] 0.3771429
$upsi #
[1] 0.1714286
$del.k.1
[1] 0.001
$b2n0
[1] 0
$b2n
[1] 1.509644e-14
$nr
[1] 3
$sing
[1] -1
$umin
[1] 0
$not.used
[1] 0
$cond.r
[1] 17672.02
$cond.h
[1] 2.001279
$scf0
[1] 1
$xnorm
[1] 1000.025
$dnorm
[1] 166.6667
$phase
[1] 0
$c.k
[1] 1
$wmax
[1] 2.2
$sig.k
[1] 1
$cfincr
[1] 2
$dirder
[1] -0.2057143
$dscal
[1] 1
$cosphi
[1] 1e-05
$violis
[1] 0
$hesstype
[1] 0
$modbfgs
[1] 0
$modnr
[1] 0
$qpterm
[1] 0
$tauqp
[1] 0
$infeas
[1] 0
$nr.update
[,1] [,2] [,3]
[1,] 1.000008 -0.005714053 -1.142493e-07
[2,] 0.000000 1.414665735 1.964581e-05
[3,] 0.000000 0.000000000 1.000000e+00
$message
[1] "KT-conditions satisfied, no further correction computed"
$runtime
[1] 0.002