带有列的数据透视表熊猫中的百分比计算

Question

我有一个数据集，其中包含来自不同供应商、地点、日期和产品的多个销售登记册。 数据集是这样的：

local   categoria   fabricante   tipo      consistencia    peso         pacote   ordem vendas_kg
AREA I  SABAO       ASATP        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 1    10
AREA I  SABAO       TEPOS        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 1    20
AREA I  SABAO       ASATP        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 1    20
AREA I  SABAO       TEPOS        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 1    30
AREA I  SABAO       ASATP        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 2    20
AREA I  SABAO       TEPOS        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 2    30
AREA I  SABAO       ASATP        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 2    20
AREA I  SABAO       TEPOS        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 2    30
AREA II SABAO       ASATP        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 1    10
AREA II SABAO       TEPOS        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 1    15
AREA II SABAO       ASATP        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 1    25
AREA II SABAO       TEPOS        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 1    35
AREA II SABAO       ASATP        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 2    20
AREA II SABAO       TEPOS        DILUIDO   LIQUIDO         1501 A 2000g PLASTICO 2    25
AREA II SABAO       TEPOS        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 2    20
AREA II SABAO       TEPOS        CAPSULA   LIQUIDO         1501 A 2000g PLASTICO 2    30

我正在使用以下代码旋转此数据集：

temp_df = pd.pivot_table(df,index=['local','tipo','ordem'], values=['vendas_kg'] , aggfunc=[np.sum], columns=['fabricante'], fill_values=0, margins=True, margins_name= 'Total')

我得到这个输出：

                                       sum           sum
                                 vendas_kg     vendas_kg
fabricante                           ASATP         TEPOS      Total
local          tipo      ordem 
AREA I         DILUIDO     1          10            20         30
                           2          20            30         50
               CAPSULA     1          10            20         30
                           2          20            30         50
AREA II        DILUIDO     1          10            15         25
                           2          20            25         45
               CAPSULA     1          25            35         55
                           2          20            30         50

我想计算每个 ['ordem'] 的百分比和每个段的百分比。 ordem、tipo 和 local，像这样：

                                       sum           sum
                                 vendas_kg     vendas_kg
fabricante                           ASATP         TEPOS     % segment  Total
local          tipo      ordem 
AREA I         DILUIDO     1          33%            66%         50%     30
                           2          40%            60%         50%     50
               CAPSULA     1          33%            66%         50%     30
                           2          40%            60%         50%     50
AREA II        DILUIDO     1          40%            60%        31.25%   25
                           2         44.44%         55.56%      47.37%   45
               CAPSULA     1         43.64%         57.36%      53.63%   55
                           2          40%            60%        53.63%   50

因此，AREA I DILUIDO 1 的总销售额为 30，ASAP 销售额占其中的 33%，TEPOS 占 66%，从 AREA I 1 的总销售额中，DILUIDOs 的销售额占 50%，依此类推。

我还想比较 ['ordem'] 之间的销售差异，例如段和 ['fabricante'] 的百分比增长，并存储在这样的新表中：

                               % change in   % change in
                                 vendas_kg     vendas_kg   % change in  % change in
fabricante                           ASATP         TEPOS     % segment        Total
AREA I  DILUIDO 1                      0             0          0                 0                 
                2                     +7%           -6%         0                20
                3                      0             0          0                 0
AREA I  CAPSULA 1                      0             0          0                 0 
                2                     +7%           -6%         0                20
                3                      0             0          0                 0
AREA II DILUIDO 1                      0             0          0                 0 
                2                    +4.44%        -4.44%    +16.12%             20
                3                      0             0          0                 0
AREA II CAPSULA 1                      0             0          0                 0  
                2                    -3.64%        +3.64%          0              5
                3                      0             0          0                 0

在过去的 5 天里，我一直坚持这一点，我在 ['fabricante']['tipo'] 和 ['local'] 中有更多的类别，所以它必须在每个类别中适用于两个以上的类别。 提前感谢您的帮助，如有疑问，请随时与我联系。

Answer 1

取百分比：

df_percent = temp_df.iloc[:, [0,1]].apply(lambda x: round(x / x.sum() * 100, 2), axis = 1)

采取变化，使用diff

df_diff_percent = df_percent.groupby(level=[0,1]).diff().fillna(0)

                    sum
                    vendas_kg
      fabricante    ASATP   TEPOS
local   tipo    ordem       
AREA I  CAPSULA 1   0.00    0.00
                2   0.00    0.00
DILUIDO         1   0.00    0.00
                2   6.67    -6.67
AREA II CAPSULA 1   0.00    0.00
                2   -41.67  41.67