資料的分群(groupby)運算¶

在這個網頁中，我們介紹一下 .groupby() 的使用。

In [ ]:

import pandas as pd
import numpy as np

import pandas as pd import numpy as np

產生虛擬資料集¶

In [ ]:

keys = np.random.choice(['A','B','C'], 20)

keys = np.random.choice(['A','B','C'], 20)

In [ ]:

keys

keys

Out[ ]:

array(['B', 'C', 'C', 'B', 'B', 'B', 'B', 'B', 'C', 'C', 'A', 'B', 'B',
       'A', 'C', 'B', 'C', 'C', 'A', 'B'], dtype='<U1')

In [ ]:

values = np.random.randint(0, 10, 20)

values = np.random.randint(0, 10, 20)

In [ ]:

df = pd.DataFrame(zip(keys, values))

df = pd.DataFrame(zip(keys, values))

In [ ]:

df

df

Out[ ]:

	0	1
0	B	4
1	C	7
2	C	5
3	B	9
4	B	7
5	B	4
6	B	9
7	B	3
8	C	5
9	C	0
10	A	1
11	B	4
12	B	2
13	A	2
14	C	1
15	B	2
16	C	6
17	C	2
18	A	5
19	B	8

In [ ]:

df.columns = ['keys','values']

df.columns = ['keys','values']

In [ ]:

df

df

Out[ ]:

	keys	values
0	B	4
1	C	7
2	C	5
3	B	9
4	B	7
5	B	4
6	B	9
7	B	3
8	C	5
9	C	0
10	A	1
11	B	4
12	B	2
13	A	2
14	C	1
15	B	2
16	C	6
17	C	2
18	A	5
19	B	8

In [ ]:

df.sort_values('keys')

df.sort_values('keys')

Out[ ]:

	keys	values
13	A	2
10	A	1
18	A	5
0	B	4
15	B	2
12	B	2
11	B	4
19	B	8
6	B	9
5	B	4
4	B	7
3	B	9
7	B	3
8	C	5
2	C	5
14	C	1
1	C	7
16	C	6
17	C	2
9	C	0

.groupby()¶

In [ ]:

grps = df.groupby('keys')

grps = df.groupby('keys')

In [ ]:

grps.get_group('A')

grps.get_group('A')

Out[ ]:

	keys	values
10	A	1
13	A	2
18	A	5

In [ ]:

grps.agg(['count', sum])

grps.agg(['count', sum])

Out[ ]:

	values
	count	sum
keys
A	3	8
B	10	52
C	7	26

In [ ]:

tuple(grps)

tuple(grps)

Out[ ]:

(('A',
     keys  values
  10    A       1
  13    A       2
  18    A       5),
 ('B',
     keys  values
  0     B       4
  3     B       9
  4     B       7
  5     B       4
  6     B       9
  7     B       3
  11    B       4
  12    B       2
  15    B       2
  19    B       8),
 ('C',
     keys  values
  1     C       7
  2     C       5
  8     C       5
  9     C       0
  14    C       1
  16    C       6
  17    C       2))

In [ ]:

dict(tuple(grps))

dict(tuple(grps))

Out[ ]:

{'A':    keys  values
 10    A       1
 13    A       2
 18    A       5,
 'B':    keys  values
 0     B       4
 3     B       9
 4     B       7
 5     B       4
 6     B       9
 7     B       3
 11    B       4
 12    B       2
 15    B       2
 19    B       8,
 'C':    keys  values
 1     C       7
 2     C       5
 8     C       5
 9     C       0
 14    C       1
 16    C       6
 17    C       2}

In [ ]:

df['keys'].unique()

df['keys'].unique()

Out[ ]:

array(['B', 'C', 'A'], dtype=object)