R 查找平均订单间隔（天数）

Question

My goal is to obtain the average number of days it takes for a given product to be purchased. If Product_A is purchased three times over a given period ('2012-12-01','2012-12-05,'2012-12-10') then our average order interval will be the average of 4 & 5 - 4.5 days.

I wrote a For Loop to calculate the interval between two points (I can use the aggregate function to calculate my mean or median by product) but I keep getting a length error. This is supposed to be a scale-able solution

Here is a sample dataframe:

product_info <- data.frame(productId = c("A", "A", "A", "B","B","B"), 
                           order_date = c("2014-05-01", "2014-05-05", "2014-05-10", "2014-06-01","2014-06-07", "2014-06-18"), stringsAsFactors=FALSE)

Here is my code:

 for (i in 2:length(unique(product_info$productId))){
  if(product_info$productId[i]==product_info$productId[i-1]){
    product_info$interval[i] <- as.integer(difftime(product_info$order_date[i],product_info$order_date[i-1]))
  }
}

My desired output should be:

product_info <- data.frame(productId = c("A", "A", "A", "B","B","B"), 
                           order_date = c("2014-05-01", "2014-05-05", "2014-05-10", "2014-06-01","2014-06-07", "2014-06-18"), 
                           interval= c("0", "4", "5", "0","6","11"), stringsAsFactors=FALSE)

Any help would be very much appreciated.

Thank you

Answer 1

You can try

  product_info$order_date <- as.Date(product_info$order_date)

  product_info$interval <- with(product_info, ave(as.numeric(order_date), 
                  productId, FUN=function(x) c(0, diff(x))))
  product_info
  productId order_date interval
1         A 2014-05-01        0
2         A 2014-05-05        4
3         A 2014-05-10        5
4         B 2014-06-01        0
5         B 2014-06-07        6
6         B 2014-06-18       11

Or using data.table

 library(data.table)#v1.9.5+
 setDT(product_info)[,interval := c(0, diff(as.Date(order_date))) , productId]

If the 'order_date' is not ordered, we have to 'order` it before doing the 'diff'

 setDT(product_info)[, order_date:= as.Date(order_date)
           ][order(order_date), interval :=as.numeric(order_date -
           shift(order_date, fill=order_date[1L])) , by = productId]
 #    productId order_date interval
 #1:         A 2014-05-01        0
 #2:         A 2014-05-05        4
 #3:         A 2014-05-10        5
 #4:         B 2014-06-01        0
 #5:         B 2014-06-07        6
 #6:         B 2014-06-18       11

Answer 2

Convert to date format -

product_info$order_date <- as.Date(product_info$order_date)

Using dplyr :

library(dplyr)
product_info %>% group_by(productId) %>%
                 mutate(interval=c(0,diff(order_date))

Answer 3

Here is a dplyr solution. You first want to convert to the date format, then ordering by the date, grouping by product and finally adding the column which is the difference between the last two days within this product. Note that the 0 days have been replaced with NA which IMHO is more applicable than 0 .

library(dplyr)
product_info <- product_info %>%
    mutate(order_date=as.Date(order_date)) %>%
    arrange(order_date) %>%
    group_by(productId) %>%
    mutate(interval=order_date-lag(order_date))

product_info
  productId order_date interval
1         A 2014-05-01  NA days
2         A 2014-05-05   4 days
3         A 2014-05-10   5 days
4         B 2014-06-01  NA days
5         B 2014-06-07   6 days
6         B 2014-06-18  11 days