Question

假设我有下一个数据框。我如何创建一个新的“ avg”列，该列是每个组的最后2个日期（“ date”）的平均值。这个想法是将其应用于具有数十万个文件的数据集，因此性能非常重要。该函数应考虑可变的月份数（例如2个月或3个月），并能够在简单平均值和中等平均值之间进行切换。

谢谢。

table1<-data.frame(group=c(1,1,1,1,2,2,2,2),date=c(201903,201902,201901,201812,201903,201902,201901,201812),price=c(10,30,50,20,2,10,9,20))


  group   date price
1     1 201903    10
2     1 201902    30
3     1 201901    50
4     1 201812    20
5     2 201903     2
6     2 201902    10
7     2 201901     9
8     2 201812    20


result<-data.frame(group=c(1,1,1,1,2,2,2,2),date=c(201903,201902,201901,201812,201903,201902,201901,201812),price=c(10,30,50,20,2,10,9,20), avg = c(20, 40, 35, NA, 6, 9.5, 14.5, NA))


  group   date price  avg
1     1 201903    10 20.0
2     1 201902    30 40.0
3     1 201901    50 35.0
4     1 201812    20   NA
5     2 201903     2  6.0
6     2 201902    10  9.5
7     2 201901     9 14.5
8     2 201812    20   NA

Answer 1

如果您的date列已排序，那么她的方法就是使用data.table：

library(data.table)
setDT(table1)[, next_price := dplyr::lead(price), by = group][, total_price := price + next_price][, avg := total_price / 2][, c("total_price", "next_price") := NULL]

table1

   group   date price  avg
1:     1 201903    10 20.0
2:     1 201902    30 40.0
3:     1 201901    50 35.0
4:     1 201812    20   NA
5:     2 201903     2  6.0
6:     2 201902    10  9.5
7:     2 201901     9 14.5
8:     2 201812    20   NA

Answer 2

首先对data.frame进行排序，以便每个组的日期递增

table1 <- table1[order(table1$group, table1$date), ]

创建带有参数月数的移动平均函数。其他功能选项可从以下位置获得：Calculating moving average

mov_avg <- function(y, months = 2){as.numeric(filter(y, rep(1 / months, months), sides = 1))}

通过此mov_avg函数使用经典的do.call-lapply-split组合

table1$avg_2months <- do.call(c, lapply(split(x=table1$price, f=table1$group), mov_avg, months=2))
table1$avg_3months <- do.call(c, lapply(split(x=table1$price, f=table1$group), mov_avg, months=3))

table1

  group   date price avg_2months avg_3months
4     1 201812    20          NA          NA
3     1 201901    50        35.0          NA
2     1 201902    30        40.0    33.33333
1     1 201903    10        20.0    30.00000
8     2 201812    20          NA          NA
7     2 201901     9        14.5          NA
6     2 201902    10         9.5    13.00000
5     2 201903     2         6.0     7.00000

R数据框按用户划分的最近几个月的平均分组

2 个答案: