如何在Python中分组、排序行数据并将其放入新列中

Question

我有这样一个数据帧：

id1   id2    c1   c2
1      a1     5     3
1      a1     6     2
1      a1     9     4
2      b1     5     2
2      b1     23    43
1      a1     34    32
3      3a     234   435

现在我想添加八个新列，如下所示：

c1_max   c2_max   c1_secondMax   c2_secondMax   c1_thirdMax c2_thirdMax   c1_min   c2_min

我想按id1和id2分组

我已经添加了c1_max c2_max c1_min c2_min列，如下所示：

features = ["c1","c2"]
maxData = all_data.groupby(['id1','id2'])[features].agg('max')
all_data = pd.merge(all_data, maxData.reset_index(), suffixes=["", "_max"], how='left', on=['id1', 'id2'])

minData= all_data.groupby(['id1','id2'])[features].agg('min')
all_data = pd.merge(all_data, minData.reset_index(), suffixes=["", "_max"], how='left', on=['id1', 'id2'])

但是，如何添加_secondMax和thirdMax。基本上，它们是这些行中的第二个最大值和第三个最大值。你知道吗

Answer 1

让我们试试这个：

smax = lambda x: x[x.rank(ascending = False) == 2]
smax.__name__ = 'secondMax'

tmax = lambda x: x[x.rank(ascending = False) == 3]
tmax.__name__ = 'thirdMax'

df_out = df.groupby(['id1','id2']).agg(['max','min',smax, tmax])
df_out.columns = [f'{i}_{j}' for i,j in df_out.columns]
df_out = df.merge(df_out, on=['id1','id2'])
df_out = df_out.where(df_out.astype(bool))
df_out

输出：

   id1 id2   c1   c2  c1_max  c1_min c1_secondMax c1_thirdMax  c2_max  c2_min c2_secondMax c2_thirdMax
0    1  a1    5    3      34       5            9           6      32       2            4           3
1    1  a1    6    2      34       5            9           6      32       2            4           3
2    1  a1    9    4      34       5            9           6      32       2            4           3
3    1  a1   34   32      34       5            9           6      32       2            4           3
4    2  b1    5    2      23       5            5         NaN      43       2            2         NaN
5    2  b1   23   43      23       5            5         NaN      43       2            2         NaN
6    3  3a  234  435     234     234          NaN         NaN     435     435          NaN         NaN