通过计算将多行转换为 pandas dataframe 中的列

Question

我有以下数据框 df

df = pd.DataFrame({'cell': ['cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1','cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1','cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2','cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2'], 'cond_one': ['cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2'], 'cond_two': ['cond_1', 'cond_2', 'cond_3', 'cond_4', 'cond_1', 'cond_2', 'cond_3', 'cond_4','cond_1', 'cond_2', 'cond_3', 'cond_4', 'cond_1', 'cond_2', 'cond_3', 'cond_4','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8'],'pvt': ['pvt1', 'pvt1', 'pvt1', 'pvt1' ,'pvt1', 'pvt1', 'pvt1', 'pvt1','pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2','pvt1', 'pvt1', 'pvt1', 'pvt1' ,'pvt1', 'pvt1', 'pvt1', 'pvt1','pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2'], 'value': ['val1', 'val2', 'val3', 'val4','val5', 'val6','val7', 'val8','val9', 'val10', 'val11', 'val12','val13', 'val14','val15', 'val16','val17', 'val18', 'val19', 'val20','val21', 'val22','val23', 'val24','val25', 'val26', 'val27', 'val28','val29', 'val30','val31', 'val32']})

cell cond_one cond_two  pvt value
cell1    cond1   cond_1 pvt1  val1
cell1    cond1   cond_2 pvt1  val2
cell1    cond1   cond_3 pvt1  val3
cell1    cond1   cond_4 pvt1  val4
cell1    cond2   cond_1 pvt1  val5
cell1    cond2   cond_2 pvt1  val6
cell1    cond2   cond_3 pvt1  val7
cell1    cond2   cond_4 pvt1  val8
cell1    cond1   cond_1 pvt2  val9
cell1    cond1   cond_2 pvt2 val10
cell1    cond1   cond_3 pvt2 val11
cell1    cond1   cond_4 pvt2 val12
cell1    cond2   cond_1 pvt2 val13
cell1    cond2   cond_2 pvt2 val14
cell1    cond2   cond_3 pvt2 val15
cell1    cond2   cond_4 pvt2 val16
cell2    cond1   cond_5 pvt1 val17
cell2    cond1   cond_6 pvt1 val18
cell2    cond1   cond_7 pvt1 val19
cell2    cond1   cond_8 pvt1 val20
cell2    cond2   cond_5 pvt1 val21
cell2    cond2   cond_6 pvt1 val22
cell2    cond2   cond_7 pvt1 val23
cell2    cond2   cond_8 pvt1 val24
cell2    cond1   cond_5 pvt2 val25
cell2    cond1   cond_6 pvt2 val26
cell2    cond1   cond_7 pvt2 val27
cell2    cond1   cond_8 pvt2 val28
cell2    cond2   cond_5 pvt2 val29
cell2    cond2   cond_6 pvt2 val30
cell2    cond2   cond_7 pvt2 val31
cell2    cond2   cond_8 pvt2 val32

列：单元格、cond_one、cond_two、pvt ：字符串类型值：float64 类型

我想根据“pvt”列中的唯一值创建列，并进行如下所示的计算。 下面是我要创建的 dataframe

df2 = pd.DataFrame({'cell': ['cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1','cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1', 'cell1','cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2','cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2', 'cell2'], 'cond_one': ['cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2','cond1', 'cond1', 'cond1', 'cond1', 'cond2', 'cond2', 'cond2', 'cond2'], 'cond_two': ['cond_1', 'cond_2', 'cond_3', 'cond_4', 'cond_1', 'cond_2', 'cond_3', 'cond_4','cond_1', 'cond_2', 'cond_3', 'cond_4', 'cond_1', 'cond_2', 'cond_3', 'cond_4','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8','cond_5', 'cond_6', 'cond_7', 'cond_8'],'pvt': ['pvt1', 'pvt1', 'pvt1', 'pvt1' ,'pvt1', 'pvt1', 'pvt1', 'pvt1','pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2','pvt1', 'pvt1', 'pvt1', 'pvt1' ,'pvt1', 'pvt1', 'pvt1', 'pvt1','pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2' ,'pvt2', 'pvt2'], 'value': ['val1', 'val2', 'val3', 'val4','val5', 'val6','val7', 'val8','val9', 'val10', 'val11', 'val12','val13', 'val14','val15', 'val16','val17', 'val18', 'val19', 'val20','val21', 'val22','val23', 'val24','val25', 'val26', 'val27', 'val28','val29', 'val30','val31', 'val32'],'pvt1': ['val1-val1', 'val2-val2', 'val3-val3', 'val4-val4' ,'val5-val5', 'val6-val6' ,'val7-val7', 'val8-val8','val9-val1', 'val10-val2', 'val11-val3', 'val12-val4' ,'val13-va5', 'val14-val6' ,'val15-val7', 'val16-val8','val17-val17', 'val18-val18', 'val19-val19', 'val20-val20' ,'val21-val21', 'val22-val22' ,'val23-val23', 'val24-val24','val25-val17', 'val26-val18', 'val27-val19', 'val28-va120' ,'val29-val21', 'val30-val22' ,'val31-val23', 'val32-val24'],'pvt2': ['val1-val9', 'val2-val10', 'val3-val11', 'val4-val12' ,'val5-val13', 'val6-val14' ,'val7-val15', 'val8-val16','val9-val9', 'val10-val10', 'val11-val11', 'val12-val12' ,'val13-val13', 'val14-val14' ,'val15-val15', 'val16-val16','val17-val25', 'val18-val26', 'val19-val27', 'val20-val28' ,'val21-val29', 'val22-val30' ,'val23-val31', 'val24-val32','val25-val25', 'val26-val26', 'val27-val27', 'val28-val28' ,'val29-val29', 'val30-val30' ,'val31-val31', 'val32-val32']})

 cell cond_one cond_two  pvt value        pvt1        pvt2
cell1    cond1   cond_1 pvt1  val1   val1-val1   val1-val9
cell1    cond1   cond_2 pvt1  val2   val2-val2  val2-val10
cell1    cond1   cond_3 pvt1  val3   val3-val3  val3-val11
cell1    cond1   cond_4 pvt1  val4   val4-val4  val4-val12
cell1    cond2   cond_1 pvt1  val5   val5-val5  val5-val13
cell1    cond2   cond_2 pvt1  val6   val6-val6  val6-val14
cell1    cond2   cond_3 pvt1  val7   val7-val7  val7-val15
cell1    cond2   cond_4 pvt1  val8   val8-val8  val8-val16
cell1    cond1   cond_1 pvt2  val9   val9-val1   val9-val9
cell1    cond1   cond_2 pvt2 val10  val10-val2 val10-val10
cell1    cond1   cond_3 pvt2 val11  val11-val3 val11-val11
cell1    cond1   cond_4 pvt2 val12  val12-val4 val12-val12
cell1    cond2   cond_1 pvt2 val13   val13-va5 val13-val13
cell1    cond2   cond_2 pvt2 val14  val14-val6 val14-val14
cell1    cond2   cond_3 pvt2 val15  val15-val7 val15-val15
cell1    cond2   cond_4 pvt2 val16  val16-val8 val16-val16
cell2    cond1   cond_5 pvt1 val17 val17-val17 val17-val25
cell2    cond1   cond_6 pvt1 val18 val18-val18 val18-val26
cell2    cond1   cond_7 pvt1 val19 val19-val19 val19-val27
cell2    cond1   cond_8 pvt1 val20 val20-val20 val20-val28
cell2    cond2   cond_5 pvt1 val21 val21-val21 val21-val29
cell2    cond2   cond_6 pvt1 val22 val22-val22 val22-val30
cell2    cond2   cond_7 pvt1 val23 val23-val23 val23-val31
cell2    cond2   cond_8 pvt1 val24 val24-val24 val24-val32
cell2    cond1   cond_5 pvt2 val25 val25-val17 val25-val25
cell2    cond1   cond_6 pvt2 val26 val26-val18 val26-val26
cell2    cond1   cond_7 pvt2 val27 val27-val19 val27-val27
cell2    cond1   cond_8 pvt2 val28 val28-va120 val28-val28
cell2    cond2   cond_5 pvt2 val29 val29-val21 val29-val29
cell2    cond2   cond_6 pvt2 val30 val30-val22 val30-val30
cell2    cond2   cond_7 pvt2 val31 val31-val23 val31-val31
cell2    cond2   cond_8 pvt2 val32 val32-val24 val32-val32

我尝试使用嵌套 for 循环来实现它，但它需要更多时间，因为我的 dataframe 有超过 10000 行。 请让我知道解决此问题的有效方法吗？

Answer 1

IIUC，您可以使用merge作为结果的起点：

pvt = (df.reset_index().merge(df, on=['cond_one', 'cond_two'], how='left')
         .assign(value=lambda x: x['value_x'] - x['value_y'],
                 column=lambda x: x.groupby('index').cumcount().add(1))
         .pivot_table(index='index', columns='column', values='value')
         .add_prefix('pvt'))
out = pd.concat([df, pvt], axis=1)

Output：

>>> out
     cell cond_one cond_two   pvt  value  pvt1  pvt2
0   cell1    cond1   cond_1  pvt1      0     0    -8
1   cell1    cond1   cond_2  pvt1      1     0    -8
2   cell1    cond1   cond_3  pvt1      2     0    -8
3   cell1    cond1   cond_4  pvt1      3     0    -8
4   cell1    cond2   cond_1  pvt1      4     0    -8
5   cell1    cond2   cond_2  pvt1      5     0    -8
6   cell1    cond2   cond_3  pvt1      6     0    -8
7   cell1    cond2   cond_4  pvt1      7     0    -8
8   cell1    cond1   cond_1  pvt2      8     8     0
9   cell1    cond1   cond_2  pvt2      9     8     0
10  cell1    cond1   cond_3  pvt2     10     8     0
11  cell1    cond1   cond_4  pvt2     11     8     0
12  cell1    cond2   cond_1  pvt2     12     8     0
13  cell1    cond2   cond_2  pvt2     13     8     0
14  cell1    cond2   cond_3  pvt2     14     8     0
15  cell1    cond2   cond_4  pvt2     15     8     0