查找排序前 n 个事务的均值

Question

让我的数据框有 2 列，customer id & transaction amount。现在对于每个唯一的客户 ID，我想找到交易金额（按降序排序）然后从排序的列中我将找到排序列表中前三笔交易的平均交易金额。

Cust_id     trans_amount
12345          100      
12345          200      
12345          170      
12345          300      
12345          250
12456          140        
12456          240       
12456          160       
12456          100          

我正在寻找的格式是，

Cust_id     trans_amount
12345          300               
12345          250      
12345          200      
12345          170      
12345          100
12456          240        
12456          160       
12456          140       
12456          100          

然后是前 3 名的平均值，即，

Cust_id    mean_for_top_3
12345         250
12456         180

中间部分，我试过了，

ddply(cust_data,.(cust_id.),summarize,sorted_amount=sort(trans_amount,,decreasing=TRUE))

但没有得到结果。请告知我如何达到我想要的输出。

Answer 1

使用data.table的解决方案：

library(data.table)
setDT(cust_data)
cust_data_sort <- cust_data[, .(trans_amount = sort(trans_amount, decreasing = TRUE)), Cust_id]
cust_data_sort[, .(mean_for_top_3 = mean(head(trans_amount, 3))), Cust_id]
   Cust_id mean_for_top_3
1:   12345            250
2:   12456            180

如果你不需要排序表cust_data_sort，那么你可以使用这个来获取平均值：

cust_data[, .(mean_for_top_3 = mean(head(sort(trans_amount, decreasing = TRUE), 3))), Cust_id]

Answer 2

使用dplyr 的惯用解决方案

df <- read.table(text = "Cust_id     trans_amount
12345          100      
                 12345          200      
                 12345          170      
                 12345          300      
                 12345          250
                 12456          140        
                 12456          240       
                 12456          160       
                 12456          100    ", header  = T)


library(dplyr)

df %>% group_by(Cust_id) %>% 
  arrange(desc(trans_amount), .by_group = T) %>%
  top_n(n = 3) %>%
  summarize(mean = mean(trans_amount))

# A tibble: 2 x 2
  Cust_id  mean
    <int> <dbl>
1   12345   250
2   12456   180

与计数的替代：

  > df %>% group_by(Cust_id) %>% 
+   #arrange(desc(trans_amount), .by_group = T) %>% 
+   mutate(count = n()) %>%
+   top_n(n = 3, wt = trans_amount) %>%
+   mutate(mean = mean(trans_amount)) %>%
+   select(Cust_id,mean,count) %>% distinct()
# A tibble: 2 x 3
# Groups:   Cust_id [2]
  Cust_id  mean count
    <int> <dbl> <int>
1   12345   250     5
2   12456   180     4
>