我有两个数据框架,第一个数据集是公司每项产品在接下来27天内预测需求的记录,如下所示:
library(tidyverse)
library(lubridate)
daily_forecast <- data.frame(
item=c("A","B","A","B"),
date_fcsted=c("2020-8-1","2020-8-1","2020-8-15","2020-8-15"),
fcsted_qty=c(100,200,200,100)
) %>%
mutate(date_fcsted=ymd(date_fcsted)) %>%
mutate(extended_date=date_fcsted+days(27))另一个数据集是每个项目的实际每日需求:
actual_orders <- data.frame(
order_date=rep(seq(ymd("2020-8-3"),ymd("2020-9-15"),by = "1 week"),2),
item=rep(c("A","B"),7),
order_qty=round(rnorm(n=14,mean=50,sd=10),0)
)我试图完成的是在第一个数据集中获取date_fcsted和extended_date中每个项目的实际总需求,然后将它们连接起来以计算预测精度。
使用tidyverse的解决方案将受到高度赞赏。
发布于 2020-10-20 01:31:08
您也可以按照@Gregor的建议尝试fuzzy_join。我添加了一个行号列,以确保您有独立于item和日期范围的唯一行(但这可能不需要)。
library(fuzzyjoin)
library(dplyr)
daily_forecast %>%
mutate(rn = row_number()) %>%
fuzzy_left_join(actual_orders,
by = c("item" = "item",
"date_fcsted" = "order_date",
"extended_date" = "order_date"),
match_fun = list(`==`, `<=`, `>=`)) %>%
group_by(rn, item.x, date_fcsted, extended_date, fcsted_qty) %>%
summarise(actual_total_demand = sum(order_qty))输出
rn item.x date_fcsted extended_date fcsted_qty actual_total_demand
<int> <chr> <date> <date> <dbl> <dbl>
1 1 A 2020-08-01 2020-08-28 100 221
2 2 B 2020-08-01 2020-08-28 200 219
3 3 A 2020-08-15 2020-09-11 200 212
4 4 B 2020-08-15 2020-09-11 100 216发布于 2020-10-20 01:05:03
您可以尝试以下方法:
library(dplyr)
daily_forecast %>%
left_join(actual_orders, by = 'item') %>%
filter(order_date >= date_fcsted & order_date <= extended_date) %>%
group_by(item, date_fcsted, extended_date, fcsted_qty) %>%
summarise(value = sum(order_qty))
# item date_fcsted extended_date fcsted_qty value
# <chr> <date> <date> <dbl> <dbl>
#1 A 2020-08-01 2020-08-28 100 179
#2 A 2020-08-15 2020-09-11 200 148
#3 B 2020-08-01 2020-08-28 200 190
#4 B 2020-08-15 2020-09-11 100 197https://stackoverflow.com/questions/64436758
复制相似问题