Dataframe内部优化

Question

我正试图在dataframe中找到一个优化解决方案。从三个单独的行中选择，我想找到一个将公式最小化的解决方案: result =(sum((B))+sum((C))/sum((A)

> set.seed(1)
> a <- sample.int(10,size = 10)
> b <- sample.int(10,size = 10)
> c <- sample.int(10,size = 10)
> df_optimize <- data.frame(a,b,c)
> df_optimize
    a  b  c
1   9  3  5
2   4  1  9
3   7  5  1
4   1  8  6
5   2  2 10
6   5  6  7
7   3 10  2
8  10  9  4
9   6  4  3
10  8  7  8
> 
> objective_function <- function(x,y,z) {
+   p_1 <- x$a + y$a + z$a
+   p_2 <- x$b + y$b + z$b
+   p_3 <- x$c + y$c + z$c
+   result <- (p_2 + p_3) / p_1
+   return(result)
+ }
> 
> #example result outputs
> objective_function(df_optimize[1,],df_optimize[3,],df_optimize[5,])
[1] 1.444444
> #and
> objective_function(df_optimize[2,],df_optimize[4,],df_optimize[6,])
[1] 3.7
> 
> # Attempt to find combination that minimizes result
> optimize(interval = df_optimize[1:10,], f = objective_function)
Error in x$a : $ operator is invalid for atomic vectors

Answer 1

这是一个野蛮的力量，网格搜索解决方案。

inx <- 1:nrow(df_optimize)
arg_space <- expand.grid(inx, inx, inx)
i <- apply(arg_space, 1, \(x) any(x[1] == x[-1] | x[2] == x[3]))
arg_space <- arg_space[!i,]

res <- sapply(1:nrow(arg_space), \(i) {
  j1 <- arg_space[i, 1, drop = TRUE]
  j2 <- arg_space[i, 2, drop = TRUE]
  j3 <- arg_space[i, 3, drop = TRUE]
  objective_function(df_optimize[j1,],df_optimize[j2,],df_optimize[j3,])
})
i_opt <- which.min(res)

arg_space[i_opt, ]
#   Var1 Var2 Var3
#29    9    3    1

res[i_opt]
#[1] 0.9545455

编辑

在缩放到更多变量时，请确保列号都与

i <- apply(arg_space, 1, \(x) length(unique(x)) == length(x))
arg_space <- arg_space[i,]

注意，这提取条件id TRUE的位置，它使用的是i而不是!i。