Question

我在R中有一个数据框，记录了许多不同品牌的客户排名偏好。数据框的示例类似于下表。实际的桌子在两个维度上都要大得多（大约80,000 x 30）。

我的表格：

+------+---------+---------+---------+---------+
| User | Brand_A | Brand_B | Brand_C | Brand_D |
+------+---------+---------+---------+---------+
| A    | 1       | NA      | 3       | 2       |
| B    | NA      | NA      | NA      | 1       |
| C    | 3       | 2       | 4       | 1       |
| D    | NA      | 1       | 2       | NA      |
+------+---------+---------+---------+---------+

其中1表示客户将品牌列为“最佳”，NA表示客户未对品牌进行排名。我想创建一个表格，为每个用户选择排名前3位（或前N位）的品牌，并输出一个可能如下所示的表格：

+------+---------+---------+---------+
| User | Ranked1 | Ranked2 | Ranked3 |
+------+---------+---------+---------+
| A    | Brand_A | Brand_D | Brand_C |
| B    | Brand_D | NA      | NA      |
| C    | Brand_D | Brand_B | Brand_A |
| D    | Brand_B | Brand_C | NA      |
+------+---------+---------+---------+

假设每个客户排名都是详尽无遗的，即如果我只使用了一个品牌，那么该品牌会自动排名为1.

我尝试使用for循环来获取所需的输出，但没有成功。我认为我缺少一些相当简单的东西。

Answer 1

一种选择是融化您的数据然后重铸它。使用data.table的此选项如下所示：

library(data.table)
dcast(setDT(melt(data, id.vars = "user"))[, rank := paste0("Ranked",value)][!is.na(value),], user ~ rank, value.var = "variable")

#  user Ranked1 Ranked2 Ranked3 Ranked4
#1    A Brand_A Brand_D Brand_C    <NA>
#2    B Brand_D    <NA>    <NA>    <NA>
#3    C Brand_D Brand_B Brand_A Brand_C
#4    D Brand_B Brand_C    <NA>    <NA>

Answer 2

您可以使用apply执行此操作：

df2=data.frame(User=df$User,t(apply(df,1,function(x) names(x)[-1][order(x[-1],na.last=NA)][1:3])))
colnames(df2)=c("User",paste0("Ranked",c(1:3)))

返回：

User Ranked1 Ranked2 Ranked3
1    A Brand_A Brand_D Brand_C
2    B Brand_D    <NA>    <NA>
3    C Brand_D Brand_B Brand_A
4    D Brand_B Brand_C    <NA>

Answer 3

与tidyverse ...

df <- read.table(header = T, text = '
User Brand_A Brand_B Brand_C Brand_D
A 1 NA 3 2
B NA NA NA 1
C 3 2 4 1
D NA 1 2 NA
')

library(tidyverse)

df %>% 
  gather(brand, rank, -User, na.rm = T) %>% 
  filter(rank < 4) %>% 
  spread(rank, brand, sep = '')

...生成

  User   rank1   rank2   rank3
1    A Brand_A Brand_D Brand_C
2    B Brand_D    <NA>    <NA>
3    C Brand_D Brand_B Brand_A
4    D Brand_B Brand_C    <NA>

R：选择每行的前N项

3 个答案: