与不同大小的data.frames匹配的字符串

Question

我有两个不同大小的 data.frame，我正在寻找最有效的方法来将字符串从一个 data.frame 匹配到另一个 data.frame，并提取一些相关信息。

这是一个例子：

两个初始 data.frames，a 和 b，以及期望的结果：

a = data.frame(term = c("red", "salad", "rope", "ball", "tent", "plane", "gift", "meat"),
               age = c(30, 24, 52, 44, 73, 44, 33, 12),
               visits = c(5, 1, 3, 2, 8, 5, 19, 3))

b = data.frame(string = c("the red ball went over the fence",
                          "sorry to see that your tent fell down",
                          "the ball fell into the red salad",
                          "serious people eat peanuts on Sundays"))

desired_result = data.frame(string = b$string,
                            num_matches = c(2, 1, 3, 0),
                            avg_age = c(37, 73, 32.66667, NA),
                            avg_visits = c(3.5, 8, 2.66667, NA))

以下是更易读格式的 data.frames：

> a
   term age visits
1   red  30      5
2 salad  24      1
3  rope  52      3
4  ball  44      2
5  tent  73      8
6 plane  44      5
7  gift  33     19
8  meat  12      3

> b
                                 string
1      the red ball went over the fence
2 sorry to see that your tent fell down
3      the ball fell into the red salad
4 serious people eat peanuts on Sundays

> desired_result
                                 string num_matches  avg_age avg_visits
1      the red ball went over the fence           2 37.00000    3.50000
2 sorry to see that your tent fell down           1 73.00000    8.00000
3      the ball fell into the red salad           3 32.66667    2.66667
4 serious people eat peanuts on Sundays           0       NA         NA

• num_matches 是“string”中“terms”的个数
• avg_age 是 "string" 中的 "terms" 的平均年龄
• avg_visits 是 "string" 中找到的 "terms" 的平均访问次数

关于如何以有效的方式实现这一点的任何想法？

谢谢。

Answer 1

你可以用base R试试这个（不需要包）：

res <- t(apply(b, 1, function(x) {
    l <- strsplit(x, " ")
    r <- unlist(lapply(unlist(l), function(y) which(a$term==y)))
    rbind(length(r), mean(a$age[r]), mean(a$visits[r]))

}))

res <- cbind(b, res)
                                 # string 1        2        3
# 1      the red ball went over the fence 2 37.00000 3.500000
# 2 sorry to see that your tent fell down 1 73.00000 8.000000
# 3      the ball fell into the red salad 3 32.66667 2.666667
# 4 serious people eat peanuts on Sundays 0      NaN      NaN

Answer 2

使用data.table，使用by = string 处理每一行。将匹配结果保存在列表中，然后按匹配结果汇总。

注意matches 列是一个列表列表，每个单元格包含一个列表。您需要使用 .() 包装匹配结果，这实际上是另一个 list()，因为 data.table 需要一个普通列的列表。

library(data.table)
library(stringr)
a = data.table(term = c("red", "salad", "rope", "ball", "tent", "plane", "gift", "meat"),
  age = c(30, 24, 52, 44, 73, 44, 33, 12),
  visits = c(5, 1, 3, 2, 8, 5, 19, 3))
b = data.table(string = c("the red ball went over the fence",
  "sorry to see that your tent fell down",
  "the ball fell into the red salad",
  "serious people eat peanuts on Sundays"))

b[, matches := vector("list", .N)]
b[, matches := .(list(str_detect(string, a[, term]))), by = string]
b[, num_matches := sum(unlist(matches)), by = string]
b[, avg_age := mean(a[unlist(matches), age]), by = string]
b[, avg_visits := mean(a[unlist(matches), visits]), by = string]

Answer 3

我会一个接一个地建立desired_result：

因此，您需要两个函数，一个用于计数，一个用于计算平均值。

首先出现：

  counter <- function(sentence, pattern)
  { 
      count <-0
      for (var in pattern)
      {
         if(grepl(pattern=var, sentence)) count <- count +1
      }
      return(count)
   }

第二个求两个平均值，两种情况都可以使用函数：

average <- function(sentence, look_up)
{ 
  pattern <- look_up[,1]
  count <-0
  summed <- 0
  for (var in pattern)
  {
    if(grepl(pattern=var, sentence)) {
      count <- count + 1
      summed <- sum(look_up[look_up[,1]==var,2]) + summed
    }
  }        
  return(summed/count)
}

您可以通过这种方式将其应用于您的数据：

先做：

desire_result <- data.frame(string = b$string)

然后获取值：

desired_result$num_match<- sapply(b$string,counter,pattern=a$term)
desired_result$avg_age<- sapply(b$string,average,look_up=a[,c(1,2)])
desired_result$avg_visit<- sapply(b$string,average,look_up=a[,c(1,3)])

这给了你现在desired_result 你的问题中提到的：

> desired_result
                                 string num_match  avg_age avg_visit
1      the red ball went over the fence         2 37.00000  3.500000
2 sorry to see that your tent fell down         1 73.00000  8.000000
3      the ball fell into the red salad         3 32.66667  2.666667
4 serious people eat peanuts on Sundays         0      NaN       NaN