如何使用 R 对销售数据进行分组？

Question

我正在尝试找到一种方法，将新产品与我拥有历史数据的产品相匹配。然后我将使用预览年份产品的历史数据来对新产品进行一些预测。

请考虑以下数据子集：

# A tibble: 13 x 11
   prdct_id prdct_grp_1 prdct_grp_2 prdct_grp_3 prdct_grp_4 Start_season January February March April sales_total
      <dbl> <chr>       <chr>       <chr>       <chr>              <dbl>   <dbl>    <dbl> <dbl> <dbl>       <dbl>
 1     1.00 WUW         SW          BH          B21                 2017    2.00    10.0   5.00  4.00        21.0
 2     2.00 WUW         SW          BK          R21                 2017    7.00     9.00  4.00  5.00        25.0
 3     3.00 MUW         NW          UW          P1                  2018    6.00     8.00 10.0   6.00        32.0
 4     4.00 LNG         KW          LW          L1                  2016    8.00     9.00 12.0   7.00        36.0
 5     5.00 QKQ         MZ          KA          AQ                  2013   10.0      8.67 16.7   8.00        43.3
 6     6.00 MUW         NW          UW          P1                  2019    0        0     0     0            0  
 7     7.00 WUW         SW          BK          R21                 2019    0        0     0     0            0  
 8     8.00 LNG         NW          UW          P2                  2014   15.1      8.67 28.7  11.0         63.4
 9     9.00 QKQ         KW          LW          L2                  2016   16.8      8.67 32.7  12.0         70.1
10    10.0  WUW         MZ          KA          AQ                  2017   18.5      8.67 36.7  13.0         76.8
11    11.0  QKQ         MZ          KA          AQ                  2019    0        0     0     0            0  
12    12.0  WUW         MZ          KA          AQ                  2019    0        0     0     0            0  
13    13.0  MUW         NW          UW          P1                  2019    0        0     0     0            0 

prdct_grp 代表产品组（例如，prdct_grp_1=WUW 表示产品在“女性内衣”中，prdct_grp_2=SW 将指定它在“泳装”组中，依此类推）。如果一个产品在同一个prdct_grp from(1-4) 中，那么我会假设他们会有非常相似的销售数字。

我希望得到以下结果

# A tibble: 3 x 11
  new_prdct_id prdct_grp_1 prdct_grp_2 prdct_grp_3 prdct_grp_4 Start_s January February March April sales_total
  <chr>        <chr>       <chr>       <chr>       <chr>         <dbl>   <dbl>    <dbl> <dbl> <dbl>       <dbl>
1 6~3          MUW         NW          UW          P1             2019    6.00     8.00 10.0   6.00        32.0
2 7~2          WUW         SW          BK          R21            2019    7.00     9.00  4.00  5.00        25.0
3 11~5         QKQ         MZ          KA          AQ             2019   10.0      9.00 17.0   8.00        43.0

我使用 tidyverse 获得了我想要的结果，但结果不是很好。

如果一个产品匹配多个产品或匹配另一个具有 2019 年开始季节的产品是另一个问题。我该如何处理？

感谢您的帮助。

最好的 一个

Answer 1

下面是一个可能的dplyr 解决方案以及详细的 cmets。请始终通过提供dput() 输出或至少提供用于创建数据集的代码 sn-p 来确保您的问题是可重现的。

# import required package
library(dplyr)

# reproduce your data frame (or at least something similar to it)
# please give more details next time
prdct_df <- data_frame(
  prdct_id = 1:13,
  prdct_grp_1 = c("WUW", "WUW", "MUW", "LNG", "QKQ", "MUW", "WUW", "LNG", "QKQ", "WUW", "QKQ", "WUW", "MUW"),
  prdct_grp_2 = c("SW", "SW", "NW", "KW", "MZ", "NW", "SW", "NW", "KW", "MZ", "MZ", "MZ", "NW"),
  prdct_grp_3 = c("BH", "BK", "UW", "LW", "KA", "UW", "BK", "UW", "LW", "KA", "KA", "KA", "UW"),
  prdct_grp_4 = c("B21", "R21", "P1", "L1", "AQ", "P1", "R21", "P2", "L2", "AQ", "AQ", "AQ", "P1"),
  Start_season = c(2017, 2017, 2018, 2016, 2013, 2019, 2019, 2014, 2016, 2017, 2019, 2019, 2019),
  January = c(2, 7, 6 , 8, 10, 0, 0, 15.1, 16.8, 18.5, 0, 0, 0),
  February = c(10, 9, 8, 9, 8.67, 0, 0, 8.86, 8.67, 8.67, 0, 0, 0),
  March = c(4, 5, 10, 12, 16.7, 0, 0, 28.7, 32.7, 36.7, 0, 0, 0),
  April  = c(4, 5, 6, 7, 8, 0, 0, 11, 12, 13, 0, 0, 0),
  sales_total = c(21, 25, 32, 36, 43.3, 0, 0, 63.4, 70.1, 76.8, 0, 0, 0)
)

# define new season in case you have additional seasons in the furture
new_prdct_seasons <- 2019 # with new seasons: c(2019, 2020, 2012) and so on

# keep the historical and new data separate (optional but clean)
# filter your data to separate new products
new_prdct_df <- prdct_df %>%
  filter(Start_season %in% new_prdct_seasons)
# filter your data to separate old products
old_prdct_df <- prdct_df %>%
  filter(!(Start_season %in% new_prdct_seasons))

# match the new and old products to get the data frame you want
final_df <- old_prdct_df %>% 
  inner_join(
    # only the first 6 columns are needed from new product data frame
    new_prdct_df[1:6], 
    # inner join by product group features
    by = c("prdct_grp_1", "prdct_grp_2", "prdct_grp_3", "prdct_grp_4")
  ) %>%
  # reorder the columns and change their names when necessary
  select(
    new_prdct_id = 12,
    old_prdct_id = 1,
    2:5,
    Start_season = 13,
    7:11
  )

# we obtained the data frame you asked for
# note that we avoided matches among new products by keeping new and old products in distinct data frames
final_df
# # A tibble: 5 x 12
#   new_prdct_id old_prdct_id prdct_grp_1 prdct_grp_2 prdct_grp_3 prdct_grp_4 Start_season January
#          <int>        <int> <chr>       <chr>       <chr>       <chr>              <dbl>   <dbl>
# 1            7            2 WUW         SW          BK          R21                 2019     7  
# 2            6            3 MUW         NW          UW          P1                  2019     6  
# 3           13            3 MUW         NW          UW          P1                  2019     6  
# 4           11            5 QKQ         MZ          KA          AQ                  2019    10  
# 5           12           10 WUW         MZ          KA          AQ                  2019    18.5
# # ... with 4 more variables: February <dbl>, March <dbl>, April <dbl>, sales_total <dbl>

# you can also exclude matches with more than one old product if needed
final_df[-3, ] # this removes the match 13-3 as there is already 6-3