R 通过id计算日期变量的连续数

Question

我有一个这样的数据库:

dat = data.frame(id = c(rep("Adam", 5), rep("Bob", 10)), 
                 card_id = c(rep("0001", 2), rep("0002", 3), rep("0003", 5), rep("0004", 5)), 
                 bill_date = c("2017-01", "2017-02", 
                               "2017-01", "2017-02", "2017-03", 
                               "2017-01", "2017-02", "2017-03", "2017-05", "2017-06", 
                               "2017-01", "2017-02", "2017-03", "2017-04", "2017-05"), stringsAsFactors = F)

#      id card_id bill_date
# 1  Adam    0001   2017-01
# 2  Adam    0001   2017-02
# 3  Adam    0002   2017-01
# 4  Adam    0002   2017-02
# 5  Adam    0002   2017-03
# 6   Bob    0003   2017-01
# 7   Bob    0003   2017-02
# 8   Bob    0003   2017-03
# 9   Bob    0003   2017-05
# 10  Bob    0003   2017-06
# 11  Bob    0004   2017-01
# 12  Bob    0004   2017-02
# 13  Bob    0004   2017-03
# 14  Bob    0004   2017-04
# 15  Bob    0004   2017-05

我想获取每个id的每个card_id的最大连续帐单数量。

预期结果:

#      id card_id max_cont_bill_num
# 1  Adam    0001                 2
# 2  Adam    0002                 3
# 3   Bob    0003                 3
# 4   Bob    0004                 5

Answer 1

我们可以尝试使用difftime和rle。创建一个函数将'bill_date'转换为Date，类，获取相邻日期的差异，转换为整数，然后获取值的run-length-id，选择最大值长度。按“id”、“card_id”分组后应用该函数

f1 <- function(x)  {
                x1 <- as.Date(paste0(x, "-01") )
                x2 <- as.integer(difftime(x1[-1], x1[-length(x1)], unit = "weeks"))
                  max(rle(c(x2[1], x2))$lengths) 
                  }

dat %>% 
   group_by(id, card_id) %>%
    summarise(max_count = f1(bill_date))        
# A tibble: 4 x 3
# Groups:   id [?]
#  id    card_id max_count
#  <chr> <chr>       <int>
#1 Adam  0001            2
#2 Adam  0002            3
#3 Bob   0003            3
#4 Bob   0004            5