Python 熊猫:groupby即将上市

Python 熊猫:groupby即将上市,python,pandas,Python,Pandas,我有如下数据: id value time 1 5 2000 1 6 2000 1 7 2000 1 5 2001 2 3 2000 2 3 2001 2 4 2005 2 5 2005 3 3 2000 3 6 2005 [[5,6,7],[5]] (this is for id 1 grouped by the id and year) [[3],[3],[4,5]] (this is for id 2

我有如下数据:

id  value   time

1   5   2000
1   6   2000
1   7   2000
1   5   2001
2   3   2000
2   3   2001
2   4   2005
2   5   2005
3   3   2000
3   6   2005
[[5,6,7],[5]] (this is for id 1 grouped by the id and year)
[[3],[3],[4,5]] (this is for id 2 grouped by the id and year)
[[3],[6]] (same logic as above)
我的最终目标是在如下列表中列出数据:

id  value   time

1   5   2000
1   6   2000
1   7   2000
1   5   2001
2   3   2000
2   3   2001
2   4   2005
2   5   2005
3   3   2000
3   6   2005
[[5,6,7],[5]] (this is for id 1 grouped by the id and year)
[[3],[3],[4,5]] (this is for id 2 grouped by the id and year)
[[3],[6]] (same logic as above)
我使用
df.groupby(['id','year'])
对数据进行了分组。但在那之后,我无法访问这些组并获取上述格式的数据

您可以使用
应用(列表)

如果您真的希望它的格式与您显示的完全相同,那么您可以按
id
分组并再次应用
list
,但这并不高效,而且这种格式可能更难使用

>>> df.groupby(['id','time'])['value'].apply(list).groupby('id').apply(list).tolist()
[[[5, 6, 7], [5]], [[3], [3], [4, 5]], [[3], [6]]]

您可以执行以下操作:

import pandas as pd

data = [[1, 5, 2000],
        [1, 6, 2000],
        [1, 7, 2000],
        [1, 5, 2001],
        [2, 3, 2000],
        [2, 3, 2001],
        [2, 4, 2005],
        [2, 5, 2005],
        [3, 3, 2000],
        [3, 6, 2005]]

df = pd.DataFrame(data=data, columns=['id', 'value', 'year'])

result = []
for name, group in df.groupby(['id']):
    result.append([g['value'].values.tolist() for _, g in group.groupby(['year'])])

for e in result:
    print(e)
df = pd.DataFrame(
    {'A': [1,1,2,2,2,2,3],
     'B':['a','b','c','d','e','f','g'],
     'C':['x','y','z','x','y','z','x']})

df.groupby('A').agg({ 'B': lambda x: list(x),'C': lambda x: list(x)})
输出

[[5, 6, 7], [5]]
[[3], [3], [4, 5]]
[[3], [6]]

如果要计算多列的列表,可以执行以下操作:

import pandas as pd

data = [[1, 5, 2000],
        [1, 6, 2000],
        [1, 7, 2000],
        [1, 5, 2001],
        [2, 3, 2000],
        [2, 3, 2001],
        [2, 4, 2005],
        [2, 5, 2005],
        [3, 3, 2000],
        [3, 6, 2005]]

df = pd.DataFrame(data=data, columns=['id', 'value', 'year'])

result = []
for name, group in df.groupby(['id']):
    result.append([g['value'].values.tolist() for _, g in group.groupby(['year'])])

for e in result:
    print(e)
df = pd.DataFrame(
    {'A': [1,1,2,2,2,2,3],
     'B':['a','b','c','d','e','f','g'],
     'C':['x','y','z','x','y','z','x']})

df.groupby('A').agg({ 'B': lambda x: list(x),'C': lambda x: list(x)})
将同时计算B和C的列表:

              B             C
A                            
1        [a, b]        [x, y]
2  [c, d, e, f]  [z, x, y, z]
3           [g]           [x]

每次我注意到自己在键入
.apply(…)
时,我都会想“你正在走出熊猫”。还有一个原因是有句谚语:“
apply()
很慢,而且没有矢量化”。然而,我必须开始习惯它。有时候这是最好的方法。