進階聚合

R 的 data.table 資料操作

Matt Dowle, Arun Srinivasan

Instructors, DataCamp

重點回顧

# Same example as seen before
## LHS := RHS Form
batrips[, c("is_dur_gt_1hour", "week_day") := 
            .(duration > 3600, wday(start_date)]

# Same as above, but in `:=`() functional form
batrips[, `:=`(is_dur_gt_1hour = duration > 3600,
               week_day = wday(start_date))]

# Update by reference with by
batrips[, n_zip_code := .N, by = zip_code]

分組一次新增多欄（參照更新）

# Functional form
batrips[, `:=`(end_dur_first = duration[1], 
               end_dur_last  = duration[.N]),
        by = end_station]

# LHS := RHS form
batrips[, c("end_dur_first", 
            "end_dur_last") := list(duration[1], duration[.N]), 
        by = end_station]

batrips[1:5]

   trip_id duration ...        end_station ... end_dur_first end_dur_last 
    139545      435 ...    Townsend at 7th ...           435          660
    139546      432 ...    Townsend at 7th ...           435          660
    139547     1523 ...    Beale at Market ...          1523          229
    139549     1620 ... Powell Street BART ...          1620          540
    139550     1617 ... Powell Street BART ...          1620          540

數值分箱

對每個 start_station 與 end_station 的唯一組合，依中位數時長：

小於 600："short"
介於 600 與 1800："medium"
否則為 "long"

j 中的多行運算式

batrips[, trip_category := {
             med_dur = median(duration, na.rm = TRUE)
             if (med_dur < 600) "short"
             else if (med_dur >= 600 & med_dur <= 1800) "medium"
             else "long"
           },
        by = .(start_station, end_station)]
batrips[1:3]

trip_id duration ... zip_code trip_category
 139545      435 ...    94612         short
 139546      432 ...    94107         short
 139547     1523 ...    94112         short

另一種作法

bin_median_duration <- function(dur) {
  med_dur <- median(dur, na.rm = TRUE)
  if (med_dur < 600) "short"
  else if (med_dur >= 600 & med_dur <= 1800) "medium"
  else "long"
}

batrips[, trip_category := bin_median_duration(duration), 
           by = .(start_station, end_station)]

整合應用：i、j、by

batrips[duration > 500, min_dur_gt_500 := min(duration), 
           by = .(start_station, end_station)]
batrips[1:3]

trip_id duration ... zip_code min_dur_gt_500
 139545      435 ...    94612             NA
 139546      432 ...    94107             NA
 139547     1523 ...    94112            502

一起來練習吧！

R 的 data.table 資料操作