搜索
简帛阁>技术文章>pandas简短介绍

pandas简短介绍

1、数据结构#

维数 名称 描述
1 Series 一维带标签单一数据类型的数组
2 DataFrame 不同数据类型的列

 

 

 

 

2、十分钟学习pandas#

      2.1、导入所需模块#

?
1
2
3
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

  2.2、创建对象#

Series

?
1
2
3
4
5
6
7
8
9
10
11
In [4]: s = pd.Series([1,3,5,np.nan,6,8])
 
In [5]: s
Out[5]:
0    1.0
1    3.0
2    5.0
3    NaN
4    6.0
5    8.0
dtype: float64

  

DataFrame

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
In [6]: dates = pd.date_range('20130101', periods=6)
 
In [7]: dates
Out[7]:
DatetimeIndex(['2013-01-01', '2013-01-02', '2013-01-03', '2013-01-04',
               '2013-01-05', '2013-01-06'],
              dtype='datetime64[ns]', freq='D')
 
In [8]: df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD')) # random.randn 生成6*4维正态随机数
 
In [9]: df
Out[9]:
                   A         B         C         D
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804
2013-01-04  0.721555 -0.706771 -1.039575  0.271860
2013-01-05 -0.424972  0.567020  0.276232 -1.087401
2013-01-06 -0.673690  0.113648 -1.478427  0.524988 

  利用字典创建DataFrame

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
In [10]: df2 = pd.DataFrame({ 'A' : 1.,
   ....:                      'B' : pd.Timestamp('20130102'),
   ....:                      'C' : pd.Series(1,index=list(range(4)),dtype='float32'),
   ....:                      'D' : np.array([3] * 4,dtype='int32'),
   ....:                      'E' : pd.Categorical(["test","train","test","train"]),
   ....:                      'F' : 'foo' })
   ....:
 
In [11]: df2
Out[11]:
     A          B    C  D      E    F
0  1.0 2013-01-02  1.0  3   test  foo
1  1.0 2013-01-02  1.0  3  train  foo
2  1.0 2013-01-02  1.0  3   test  foo
3  1.0 2013-01-02  1.0  3  train  foo

  每一列都有不同的类型

?
1
2
3
4
5
6
7
8
9
In [12]: df2.dtypes
Out[12]:
A           float64
B    datetime64[ns]
C           float32
D             int32
E          category
F            object
dtype: object

  2.3查看数据#

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
In [14]: df.head() #查看数据开头几行,默认5行
Out[14]:
                   A         B         C         D
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804
2013-01-04  0.721555 -0.706771 -1.039575  0.271860
2013-01-05 -0.424972  0.567020  0.276232 -1.087401
 
In [15]: df.tail(3) #查看数据结尾3行
Out[15]:
                   A         B         C         D
2013-01-04  0.721555 -0.706771 -1.039575  0.271860
2013-01-05 -0.424972  0.567020  0.276232 -1.087401
2013-01-06 -0.673690  0.113648 -1.478427  0.524988

  单独查看数据的索引、列名、值

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
In [16]: df.index
Out[16]:
DatetimeIndex(['2013-01-01', '2013-01-02', '2013-01-03', '2013-01-04',
               '2013-01-05', '2013-01-06'],
              dtype='datetime64[ns]', freq='D')
 
In [17]: df.columns
Out[17]: Index(['A', 'B', 'C', 'D'], dtype='object')
 
In [18]: df.values
Out[18]:
array([[ 0.4691, -0.2829, -1.5091, -1.1356],
       [ 1.2121, -0.17320.1192, -1.0442],
       [-0.8618, -2.1046, -0.49491.0718],
       [ 0.7216, -0.7068, -1.03960.2719],
       [-0.425 0.567 0.2762, -1.0874],
       [-0.67370.1136, -1.47840.525 ]])

  describe()显示数据统计概要

?
1
2
3
4
5
6
7
8
9
10
11
In [19]: df.describe()
Out[19]:
              A         B         C         D
count  6.000000  6.000000  6.000000  6.000000
mean   0.073711 -0.431125 -0.687758 -0.233103
std    0.843157  0.922818  0.779887  0.973118
min   -0.861849 -2.104569 -1.509059 -1.135632
25%   -0.611510 -0.600794 -1.368714 -1.076610
50%    0.022070 -0.228039 -0.767252 -0.386188
75%    0.658444  0.041933 -0.034326  0.461706
max    1.212112  0.567020  0.276232  1.071804

  转置数据

?
1
2
3
4
5
6
7
In [20]: df.T
Out[20]:
   2013-01-01  2013-01-02  2013-01-03  2013-01-04  2013-01-05  2013-01-06
A    0.469112    1.212112   -0.861849    0.721555   -0.424972   -0.673690
B   -0.282863   -0.173215   -2.104569   -0.706771    0.567020    0.113648
C   -1.509059    0.119209   -0.494929   -1.039575    0.276232   -1.478427
D   -1.135632   -1.044236    1.071804    0.271860   -1.087401    0.524988

  按index排序

?
1
2
3
4
5
6
7
8
9
In [21]: df.sort_index(axis=1, ascending=False) # axis=0排序行,ascending=F 递减
Out[21]:
                   D         C         B         A
2013-01-01 -1.135632 -1.509059 -0.282863  0.469112
2013-01-02 -1.044236  0.119209 -0.173215  1.212112
2013-01-03  1.071804 -0.494929 -2.104569 -0.861849
2013-01-04  0.271860 -1.039575 -0.706771  0.721555
2013-01-05 -1.087401  0.276232  0.567020 -0.424972
2013-01-06  0.524988 -1.478427  0.113648 -0.673690

  按值排序

?
1
2
3
4
5
6
7
8
9
In [22]: df.sort_values(by='B') #根据B值排序, 默认ascending =T 递增
Out[22]:
                   A         B         C         D
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804
2013-01-04  0.721555 -0.706771 -1.039575  0.271860
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-06 -0.673690  0.113648 -1.478427  0.524988
2013-01-05 -0.424972  0.567020  0.276232 -1.087401

  2.4选择数据#

       2.4.1通过标签选择

         通过列名选择列

?
1
2
3
4
5
6
7
8
9
In [23]: df['A']
Out[23]:
2013-01-01    0.469112
2013-01-02    1.212112
2013-01-03   -0.861849
2013-01-04    0.721555
2013-01-05   -0.424972
2013-01-06   -0.673690
Freq: D, Name: A, dtype: float64

  通过 [ ] 选择行

?
1
2
3
4
5
6
7
8
9
10
11
12
13
In [24]: df[0:3]
Out[24]:
                   A         B         C         D
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804
 
In [25]: df['20130102':'20130104']
Out[25]:
                   A         B         C         D
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804
2013-01-04  0.721555 -0.706771 -1.039575  0.271860

  根据标签选择

?
1
2
3
4
5
6
7
In [26]: df.loc[dates[0]] # 等于df.loc["2013-01-01"]
Out[26]:
A    0.469112
B   -0.282863
C   -1.509059
D   -1.135632
Name: 2013-01-01 00:00:00, dtype: float64

  根据多轴标签选择

?
1
2
3
4
5
6
7
8
9
In [27]: df.loc[:,['A','B']]
Out[27]:
                   A         B
2013-01-01  0.469112 -0.282863
2013-01-02  1.212112 -0.173215
2013-01-03 -0.861849 -2.104569
2013-01-04  0.721555 -0.706771
2013-01-05 -0.424972  0.567020
2013-01-06 -0.673690  0.113648

  根据行和列选择

?
1
2
3
4
5
6
In [28]: df.loc['20130102':'20130104',['A','B']]
Out[28]:
                   A         B
2013-01-02  1.212112 -0.173215
2013-01-03 -0.861849 -2.104569
2013-01-04  0.721555 -0.706771

  降维

?
1
2
3
4
5
6
7
8
9
10
11
In [29]: df.loc['20130102',['A','B']] #降成一维
Out[29]:
A    1.212112
B   -0.173215
Name: 2013-01-02 00:00:00, dtype: float64
 
In [30]: df.loc[dates[0],'A'] #降成标量
Out[30]: 0.46911229990718628
 
In [31]: df.at[dates[0],'A'] #等于上
Out[31]: 0.46911229990718628

  2.4.2 通过整数的位置坐标选择

?
1
2
3
4
5
6
7
In [32]: df.iloc[3] # 选择了第3行
Out[32]:
A    0.721555
B   -0.706771
C   -1.039575
D    0.271860
Name: 2013-01-04 00:00:00, dtype: float64

  通过位置坐标切片

?
1
2
3
4
5
In [33]: df.iloc[3:5,0:2] # 第3-5行的0-2列,不包括5和2
Out[33]:
                   A         B
2013-01-04  0.721555 -0.706771
2013-01-05 -0.424972  0.567020

  通过整数坐标列表选择

?
1
2
3
4
5
6
In [34]: df.iloc[[1,2,4],[0,2]]
Out[34]:
                   A         C
2013-01-02  1.212112  0.119209
2013-01-03 -0.861849 -0.494929
2013-01-05 -0.424972  0.276232

  2.4.3 布尔型选择

       通过列值大小索引

?
1
2
3
4
5
6
In [39]: df[df.A > 0] #列A大于0的行选出来
Out[39]:
                   A         B         C         D
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632
2013-01-02  1.212112 -0.173215  0.119209 -1.044236
2013-01-04  0.721555 -0.706771 -1.039575  0.271860

  整个DateFrame 选择

?
1
2
3
4
5
6
7
8
9
In [40]: df[df > 0] #不符合条件的变成了NAN
Out[40]:
                   A         B         C         D
2013-01-01  0.469112       NaN       NaN       NaN
2013-01-02  1.212112       NaN  0.119209       NaN
2013-01-03       NaN       NaN       NaN  1.071804
2013-01-04  0.721555       NaN       NaN  0.271860
2013-01-05       NaN  0.567020  0.276232       NaN
2013-01-06       NaN  0.113648       NaN  0.524988

  利用isin() 过滤选择

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In [41]: df2 = df.copy()
 
In [42]: df2['E'] = ['one', 'one','two','three','four','three']
 
In [43]: df2
Out[43]:
                   A         B         C         D      E
2013-01-01  0.469112 -0.282863 -1.509059 -1.135632    one
2013-01-02  1.212112 -0.173215  0.119209 -1.044236    one
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804    two
2013-01-04  0.721555 -0.706771 -1.039575  0.271860  three
2013-01-05 -0.424972  0.567020  0.276232 -1.087401   four
2013-01-06 -0.673690  0.113648 -1.478427  0.524988  three
 
In [44]: df2[df2['E'].isin(['two','four'])]
Out[44]:
                   A         B         C         D     E
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804   two
2013-01-05 -0.424972  0.567020  0.276232 -1.087401  four
 
In []: df2['E'].isin(['two','four']) #生成的是一个布尔数组
Out[]:
2013-01-01    False
2013-01-02    False
2013-01-03     True
2013-01-04    False
2013-01-05     True
2013-01-06    False
Freq: D, Name: E, dtype: bool
 
In[]: df2[[False,False,True,False,True,False]] #和第44步效果一样
Out[]:
                   A         B         C         D     E
2013-01-03 -0.861849 -2.104569 -0.494929  1.071804   two
2013-01-05 -0.424972  0.567020  0.276232 -1.087401  four

 2.5 修改数据 #

通过索引自动把列组合进数据

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
In [45]: s1 = pd.Series([1,2,3,4,5,6], index=pd.date_range('20130102', periods=6)) #索引和 dateframe 一样
 
In [46]: s1
Out[46]:
2013-01-02    1
2013-01-03    2
2013-01-04    3
2013-01-05    4
2013-01-06    5
2013-01-07    6
Freq: D, dtype: int64
 
In [47]: df['F'] = s1 #相同索引自动组合
In [48]: df.at[dates[0],'A'] = 0 #更改0,“A”这个位置的值为0
In [49]: df.iat[0,1] = 0  #更改0,1位置的值为0
In [50]: df.loc[:,'D'] = np.array([5] * len(df)) #把D列全改为5
In [51]: df
Out[51]:
                   A         B         C  D    F
2013-01-01  0.000000  0.000000 -1.509059  5  NaN
2013-01-02  1.212112 -0.173215  0.119209  5  1.0
2013-01-03 -0.861849 -2.104569 -0.494929  5  2.0
2013-01-04  0.721555 -0.706771 -1.039575  5  3.0
2013-01-05 -0.424972  0.567020  0.276232  5  4.0
2013-01-06 -0.673690  0.113648 -1.478427  5  5.0

  原表按位置覆盖数据

?
1
2
3
4
5
6
7
8
9
10
11
12
13
In [52]: df2 = df.copy()
 
In [53]: df2[df2 > 0] = -df2 #只覆盖非空值,即大于0的值
 
In [54]: df2
Out[54]:
                   A         B         C  D    F
2013-01-01  0.000000  0.000000 -1.509059 -5  NaN
2013-01-02 -1.212112 -0.173215 -0.119209 -5 -1.0
2013-01-03 -0.861849 -2.104569 -0.494929 -5 -2.0
2013-01-04 -0.721555 -0.706771 -1.039575 -5 -3.0
2013-01-05 -0.424972 -0.567020 -0.276232 -5 -4.0
2013-01-06 -0.673690 -0.113648 -1.478427 -5 -5.0

  2.6处理空值#

         reindex()允许修改、删除、增加 index 。

?
1
2
3
4
5
6
7
8
9
10
11
In [55]: df1 = df.reindex(index=dates[0:4], columns=list(df.columns) + ['E'])
 
In [56]: df1.loc[dates[0]:dates[1],'E'] = 1  #虽然能用索引改值,但是df1.index 得出依然是 2013-01-01 等日期
 
In [57]: df1
Out[57]:
                   A         B         C  D    F    E
2013-01-01  0.000000  0.000000 -1.509059  5  NaN  1.0
2013-01-02  1.212112 -0.173215  0.119209  5  1.0  1.0
2013-01-03 -0.861849 -2.104569 -0.494929  5  2.0  NaN
2013-01-04  0.721555 -0.706771 -1.039575  5  3.0  NaN

  删除带有空值得行

?
1
2
3
4
In [58]: df1.dropna(how='any')
Out[58]:
                   A         B         C  D    F    E
2013-01-02  1.212112 -0.173215  0.119209  5  1.0  1.0

  填充空值

?
1
2
3
4
5
6
7
In [59]: df1.fillna(value=5)
Out[59]:
                   A         B         C  D    F    E
2013-01-01  0.000000  0.000000 -1.509059  5  5.0  1.0
2013-01-02  1.212112 -0.173215  0.119209  5  1.0  1.0
2013-01-03 -0.861849 -2.104569 -0.494929  5  2.0  5.0
2013-01-04  0.721555 -0.706771 -1.039575  5  3.0  5.0

  是nan就显示ture

?
1
2
3
4
5
6
7
In [60]: pd.isna(df1)
Out[60]:
                A      B      C      D      F      E
2013-01-01  False  False  False  False   True  False
2013-01-02  False  False  False  False  False  False
2013-01-03  False  False  False  False  False   True
2013-01-04  False  False  False  False  False   True

  2.6 操作#

         2.6.1 统计

均值

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
In [61]: df.mean() #各列的均值
Out[61]:
A   -0.004474
B   -0.383981
C   -0.687758
D    5.000000
F    3.000000
dtype: float64
 
In [62]: df.mean(1) #各行的均值
Out[62]:
2013-01-01    0.872735
2013-01-02    1.431621
2013-01-03    0.707731
2013-01-04    1.395042
2013-01-05    1.883656
2013-01-06    1.592306
Freq: D, dtype: float64

  shift() 索引不变数据下移个数,sub() 减法

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
In [63]: s = pd.Series([1,3,5,np.nan,6,8], index=dates).shift(2) # 原数据是 1,3,5,nan, 6,8 ,下移2
 
In [64]: s
Out[64]:
2013-01-01    NaN
2013-01-02    NaN
2013-01-03    1.0
2013-01-04    3.0
2013-01-05    5.0
2013-01-06    NaN
Freq: D, dtype: float64
 
In [65]: df.sub(s, axis='index') #每一列都减去s , 值-NAN = NAN
Out[65]:
                   A         B         C    D    F
2013-01-01       NaN       NaN       NaN  NaN  NaN
2013-01-02       NaN       NaN       NaN  NaN  NaN
2013-01-03 -1.861849 -3.104569 -1.494929  4.0  1.0
2013-01-04 -2.278445 -3.706771 -4.039575  2.0  0.0
2013-01-05 -5.424972 -4.432980 -4.723768  0.0 -1.0
2013-01-06       NaN       NaN       NaN  NaN  NaN

  2.6.2 应用 (Apply)

       对数据应用函数

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
In [66]: df.apply(np.cumsum) #cumsum() 累积加和
Out[66]:
                   A         B         C   D     F
2013-01-01  0.000000  0.000000 -1.509059   5   NaN
2013-01-02  1.212112 -0.173215 -1.389850  10   1.0
2013-01-03  0.350263 -2.277784 -1.884779  15   3.0
2013-01-04  1.071818 -2.984555 -2.924354  20   6.0
2013-01-05  0.646846 -2.417535 -2.648122  25  10.0
2013-01-06 -0.026844 -2.303886 -4.126549  30  15.0
 
In [67]: df.apply(lambda x: x.max() - x.min())
Out[67]:
A    2.073961
B    2.671590
C    1.785291
D    0.000000
F    4.000000
dtype: float64

  2.6.3 直方图

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
In [68]: s = pd.Series(np.random.randint(0, 7, size=10))
 
In [69]: s
Out[69]:
0    4
1    2
2    1
3    2
4    6
5    4
6    4
7    6
8    4
9    4
dtype: int64
 
In [70]: s.value_counts()
Out[70]:
4    5
6    2
2    2
1    1
dtype: int64

  2.6.4 字符串操作

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
In [71]: s = pd.Series(['A', 'B', 'C', 'Aaba', 'Baca', np.nan, 'CABA', 'dog', 'cat'])
 
In [72]: s.str.lower() #str.lower() 小写
Out[72]:
0       a
1       b
2       c
3    aaba
4    baca
5     NaN
6    caba
7     dog
8     cat
dtype: object

  2.7 合并 (Merge)#

        合并多个数组(concat)

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
In [73]: df = pd.DataFrame(np.random.randn(10, 4))
 
In [74]: df
Out[74]:
          0         1         2         3
0 -0.548702  1.467327 -1.015962 -0.483075
1  1.637550 -1.217659 -0.291519 -1.745505
2 -0.263952  0.991460 -0.919069  0.266046
3 -0.709661  1.669052  1.037882 -1.705775
4 -0.919854 -0.042379  1.247642 -0.009920
5  0.290213  0.495767  0.362949  1.548106
6 -1.131345 -0.089329  0.337863 -0.945867
7 -0.932132  1.956030  0.017587 -0.016692
8 -0.575247  0.254161 -1.143704  0.215897
9  1.193555 -0.077118 -0.408530 -0.862495
 
# break it into pieces
In [75]: pieces = [df[:3], df[3:7], df[7:]]
In []    : pieces
Out[]   :    #结构类似于这样,随机数变了
[          0         1         2         3
 0  0.297766 -0.983514  0.813754 -0.404223
 1  1.274169  0.317278 -0.786275  0.511142
 2 -0.755956  0.411957  1.441029  0.909508,
           0         1         2         3
 3 -0.996374 -0.284406  0.307254 -0.192736
 4  0.350560  0.561135 -1.193311  0.444860
 5  3.098968  1.080112 -0.823096  1.091375
 6  1.576157  1.060650 -0.398183 -0.438593,
           0         1         2         3
 7 -0.300971  0.475688 -2.622464  0.785068
 8 -1.335871 -1.571492  2.505800  1.000592
 9  0.522409 -0.792110  0.746916 -2.439753]
 
In [76]: pd.concat(pieces)
Out[76]:
          0         1         2         3
0 -0.548702  1.467327 -1.015962 -0.483075
1  1.637550 -1.217659 -0.291519 -1.745505
2 -0.263952  0.991460 -0.919069  0.266046
3 -0.709661  1.669052  1.037882 -1.705775
4 -0.919854 -0.042379  1.247642 -0.009920
5  0.290213  0.495767  0.362949  1.548106
6 -1.131345 -0.089329  0.337863 -0.945867
7 -0.932132  1.956030  0.017587 -0.016692
8 -0.575247  0.254161 -1.143704  0.215897
9  1.193555 -0.077118 -0.408530 -0.862495

  根据数据连接(jion)

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
In [77]: left = pd.DataFrame({'key': ['foo', 'foo'], 'lval': [1, 2]})
 
In [78]: right = pd.DataFrame({'key': ['foo', 'foo'], 'rval': [4, 5]})
 
In [79]: left
Out[79]:
   key  lval
0  foo     1
1  foo     2
 
In [80]: right
Out[80]:
   key  rval
0  foo     4
1  foo     5
 
In [81]: pd.merge(left, right, on='key')
Out[81]:
   key  lval  rval
0  foo     1     4
1  foo     1     5
2  foo     2     4
3  foo     2     5
 
 
In [82]: left = pd.DataFrame({'key': ['foo', 'bar'], 'lval': [1, 2]})
 
In [83]: right = pd.DataFrame({'key': ['foo', 'bar'], 'rval': [4, 5]})
 
In [84]: left
Out[84]:
   key  lval
0  foo     1
1  bar     2
 
In [85]: right
Out[85]:
   key  rval
0  foo     4
1  bar     5
 
In [86]: pd.merge(left, right, on='key')
Out[86]:
   key  lval  rval
0  foo     1     4
1  bar     2     5

  贴上(Append)

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
In [87]: df = pd.DataFrame(np.random.randn(8, 4), columns=['A','B','C','D'])
 
In [88]: df
Out[88]:
          A         B         C         D
0  1.346061  1.511763  1.627081 -0.990582
1 -0.441652  1.211526  0.268520  0.024580
2 -1.577585  0.396823 -0.105381 -0.532532
3  1.453749  1.208843 -0.080952 -0.264610
4 -0.727965 -0.589346  0.339969 -0.693205
5 -0.339355  0.593616  0.884345  1.591431
6  0.141809  0.220390  0.435589  0.192451
7 -0.096701  0.803351  1.715071 -0.708758
 
In [89]: s = df.iloc[3] #注意是第3行,而不是第2行,也是按0算的
 
In [90]: df.append(s, ignore_index=True)
Out[90]:
          A         B         C         D
0  1.346061  1.511763  1.627081 -0.990582
1 -0.441652  1.211526  0.268520  0.024580
2 -1.577585  0.396823 -0.105381 -0.532532
3  1.453749  1.208843 -0.080952 -0.264610
4 -0.727965 -0.589346  0.339969 -0.693205
5 -0.339355  0.593616  0.884345  1.591431
6  0.141809  0.220390  0.435589  0.192451
7 -0.096701  0.803351  1.715071 -0.708758
8  1.453749  1.208843 -0.080952 -0.264610

  2.8 组合(Grouping)#

       1、把数据按某些标准分裂成组

       2、每组单独应用函数

       3、把结果组合到一起

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
In [91]: df = pd.DataFrame({'A' : ['foo', 'bar', 'foo', 'bar',
   ....:                           'foo', 'bar', 'foo', 'foo'],
   ....:                    'B' : ['one', 'one', 'two', 'three',
   ....:                           'two', 'two', 'one', 'three'],
   ....:                    'C' : np.random.randn(8),
   ....:                    'D' : np.random.randn(8)})
   ....:
 
In [92]: df
Out[92]:
     A      B         C         D
0  foo    one -1.202872 -0.055224
1  bar    one -1.814470  2.395985
2  foo    two  1.018601  1.552825
3  bar  three -0.595447  0.166599
4  foo    two  1.395433  0.047609
5  bar    two -0.392670 -0.136473
6  foo    one  0.007207 -0.561757
7  foo  three  1.928123 -1.623033

  分组并用sum()函数

?
1
2
3
4
5
6
In [93]: df.groupby('A').sum()
Out[93]:
            C        D
A                    
bar -2.802588  2.42611
foo  3.146492 -0.63958

  多列分组,求sum()

?
1
2
3
4
5
6
7
8
9
10
In [94]: df.groupby(['A','B']).sum()
Out[94]:
                  C         D
A   B                       
bar one   -1.814470  2.395985
    three -0.595447  0.166599
    two   -0.392670 -0.136473
foo one   -1.195665 -0.616981
    three  1.928123 -1.623033
    two    2.414034  1.600434

  分组顺序不同,结果不同

?
1
>>> df.groupby(['B','A']).sum()

  

        2.9 更改形状(reshaping)#

          2.9.1 堆叠

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
In [95]: tuples = list(zip(*[['bar', 'bar', 'baz', 'baz',
   ....:                      'foo', 'foo', 'qux', 'qux'],
   ....:                     ['one', 'two', 'one', 'two',
   ....:                      'one', 'two', 'one', 'two']]))
   ....:
Out[]  :
[('bar', 'one'),
 ('bar', 'two'),
 ('baz', 'one'),
 ('baz', 'two'),
 ('foo', 'one'),
 ('foo', 'two'),
 ('qux', 'one'),
 ('qux', 'two')]
 
In [96]: index = pd.MultiIndex.from_tuples(tuples, names=['first', 'second'])
 
In [97]: df = pd.DataFrame(np.random.randn(8, 2), index=index, columns=['A', 'B'])
In [98]: df2 = df[:4]
 
In [99]: df2
Out[99]:
                     A         B
first second                   
bar   one     0.029399 -0.542108
      two     0.282696 -0.087302
baz   one    -1.575170  1.771208
      two     0.816482  1.100230

  stack() 将两列压缩到一列

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
In [100]: stacked = df2.stack()
 
In [101]: stacked
Out[101]:
first  second  
bar    one     A    0.029399
               B   -0.542108
       two     A    0.282696
               B   -0.087302
baz    one     A   -1.575170
               B    1.771208
       two     A    0.816482
               B    1.100230
dtype: float64

  2.9.2 反堆叠(unstack)

默认是最后一级

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
In [102]: stacked.unstack()
Out[102]:
                     A         B
first second                   
bar   one     0.029399 -0.542108
      two     0.282696 -0.087302
baz   one    -1.575170  1.771208
      two     0.816482  1.100230
 
In [103]: stacked.unstack(1)
Out[103]:
second        one       two
first                     
bar   A  0.029399  0.282696
      B -0.542108 -0.087302
baz   A -1.575170  0.816482
      1.771208  1.100230
 
In [104]: stacked.unstack(0)
Out[104]:
first          bar       baz
second                     
one    A  0.029399 -1.575170
       B -0.542108  1.771208
two    A  0.282696  0.816482
       B -0.087302  1.100230

  2.9.4 数据透视表

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
In [105]: df = pd.DataFrame({'A' : ['one', 'one', 'two', 'three'] * 3,
   .....:                    'B' : ['A', 'B', 'C'] * 4,
   .....:                    'C' : ['foo', 'foo', 'foo', 'bar', 'bar', 'bar'] * 2,
   .....:                    'D' : np.random.randn(12),
   .....:                    'E' : np.random.randn(12)})
   .....:
 
In [106]: df
Out[106]:
        A  B    C         D         E
0     one  A  foo  1.418757 -0.179666
1     one  B  foo -1.879024  1.291836
2     two  C  foo  0.536826 -0.009614
3   three  A  bar  1.006160  0.392149
4     one  B  bar -0.029716  0.264599
5     one  C  bar -1.146178 -0.057409
6     two  A  foo  0.100900 -1.425638
7   three  B  foo -1.035018  1.024098
8     one  C  foo  0.314665 -0.106062
9     one  A  bar -0.773723  1.824375
10    two  B  bar -1.170653  0.595974
11  three  C  bar  0.648740  1.167115
 
In [107]: pd.pivot_table(df, values='D', index=['A', 'B'], columns=['C'])
Out[107]:
C             bar       foo
A     B                   
one   A -0.773723  1.418757
      B -0.029716 -1.879024
      C -1.146178  0.314665
three A  1.006160       NaN
      B       NaN -1.035018
      0.648740       NaN
two   A       NaN  0.100900
      B -1.170653       NaN
      C       NaN  0.536826

  2.10 时间Series#

?
1
2
3
4
5
6
7
8
In [108]: rng = pd.date_range('1/1/2012', periods=100, freq='S')
 
In [109]: ts = pd.Series(np.random.randint(0, 500, len(rng)), index=rng)
 
In [110]: ts.resample('5Min').sum()
Out[110]:
2012-01-01    25083
Freq: 5T, dtype: int64

  时区表示

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
In [111]: rng = pd.date_range('3/6/2012 00:00', periods=5, freq='D')
 
In [112]: ts = pd.Series(np.random.randn(len(rng)), rng)
 
In [113]: ts
Out[113]:
2012-03-06    0.464000
2012-03-07    0.227371
2012-03-08   -0.496922
2012-03-09    0.306389
2012-03-10   -2.290613
Freq: D, dtype: float64
 
In [114]: ts_utc = ts.tz_localize('UTC')
 
In [115]: ts_utc
Out[115]:
2012-03-06 00:00:00+00:00    0.464000
2012-03-07 00:00:00+00:00    0.227371
2012-03-08 00:00:00+00:00   -0.496922
2012-03-09 00:00:00+00:00    0.306389
2012-03-10 00:00:00+00:00   -2.290613
Freq: D, dtype: float64

  转换到另一时区

?
1
2
3
4
5
6
7
8
In [116]: ts_utc.tz_convert('US/Eastern')
Out[116]:
2012-03-05 19:00:00-05:00    0.464000
2012-03-06 19:00:00-05:00    0.227371
2012-03-07 19:00:00-05:00   -0.496922
2012-03-08 19:00:00-05:00    0.306389
2012-03-09 19:00:00-05:00   -2.290613
Freq: D, dtype: float64

  时间跨度转换

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
In [117]: rng = pd.date_range('1/1/2012', periods=5, freq='M')
 
In [118]: ts = pd.Series(np.random.randn(len(rng)), index=rng)
 
In [119]: ts
Out[119]:
2012-01-31   -1.134623
2012-02-29   -1.561819
2012-03-31   -0.260838
2012-04-30    0.281957
2012-05-31    1.523962
Freq: M, dtype: float64
 
In [120]: ps = ts.to_period()
 
In [121]: ps
Out[121]:
2012-01   -1.134623
2012-02   -1.561819
2012-03   -0.260838
2012-04    0.281957
2012-05    1.523962
Freq: M, dtype: float64
 
In [122]: ps.to_timestamp()
Out[122]:
2012-01-01   -1.134623
2012-02-01   -1.561819
2012-03-01   -0.260838
2012-04-01    0.281957
2012-05-01    1.523962
Freq: MS, dtype: float64

  时间戳转时期

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
In [123]: prng = pd.period_range('1990Q1', '2000Q4', freq='Q-NOV')
 
In [124]: ts = pd.Series(np.random.randn(len(prng)), prng)
 
In [125]: ts.index = (prng.asfreq('M', 'e') + 1).asfreq('H', 's') + 9
 
In [126]: ts.head()
Out[126]:
1990-03-01 09:00   -0.902937
1990-06-01 09:00    0.068159
1990-09-01 09:00   -0.057873
1990-12-01 09:00   -0.368204
1991-03-01 09:00   -1.144073
Freq: H, dtype: float64

  2.11 分类(categorical)#

?
1
2
3
4
5
6
7
8
9
10
11
12
13
In [127]: df = pd.DataFrame({"id":[1,2,3,4,5,6], "raw_grade":['a', 'b', 'b', 'a', 'a', 'e']})
In [128]: df["grade"] = df["raw_grade"].astype("category")
 
In [129]: df["grade"]
Out[129]:
0    a
1    b
2    b
3    a
4    a
5    e
Name: grade, dtype: category
Categories (3, object): [a, b, e]

  重命名

?
1
In [130]: df["grade"].cat.categories = ["very good", "good", "very bad"]

  重新排列类别,同时添加没有值的类别

?
1
2
3
4
5
6
7
8
9
10
11
12
In [131]: df["grade"] = df["grade"].cat.set_categories(["very bad", "bad", "medium", "good", "very good"])
 
In [132]: df["grade"]
Out[132]:
0    very good
1         good
2         good
3    very good
4    very good
5     very bad
Name: grade, dtype: category
Categories (5, object): [very bad, bad, medium, good, very good]

  根据类别排序,而不是字符

?
1
2
3
4
5
6
7
8
9
In [133]: df.sort_values(by="grade")
Out[133]:
   id raw_grade      grade
5   6         e   very bad
1   2         b       good
2   3         b       good
0   1         a  very good
3   4         a  very good
4   5         a  very good

  根据类别分组,同时显示没值得类别

?
1
2
3
4
5
6
7
8
9
In [134]: df.groupby("grade").size()
Out[134]:
grade
very bad     1
bad          0
medium       0
good         2
very good    3
dtype: int64

  2.12 输入输出数据#

         2.12.1 CSV

?
1
In [141]: df.to_csv('foo.csv')

  读入

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [142]: pd.read_csv('foo.csv')
Out[142]:
     Unnamed: 0          A          B         C          D
0    2000-01-01   0.266457  -0.399641 -0.219582   1.186860
1    2000-01-02  -1.170732  -0.345873  1.653061  -0.282953
2    2000-01-03  -1.734933   0.530468  2.060811  -0.515536
3    2000-01-04  -1.555121   1.452620  0.239859  -1.156896
4    2000-01-05   0.578117   0.511371  0.103552  -2.428202
5    2000-01-06   0.478344   0.449933 -0.741620  -1.962409
6    2000-01-07   1.235339  -0.091757 -1.543861  -1.084753
..          ...        ...        ...       ...        ...
993  2002-09-20 -10.628548  -9.153563 -7.883146  28.313940
994  2002-09-21 -10.390377  -8.727491 -6.399645  30.914107
995  2002-09-22  -8.985362  -8.485624 -4.669462  31.367740
996  2002-09-23  -9.558560  -8.781216 -4.499815  30.518439
997  2002-09-24  -9.902058  -9.340490 -4.386639  30.105593
998  2002-09-25 -10.216020  -9.480682 -3.933802  29.758560
999  2002-09-26 -11.856774 -10.671012 -3.216025  29.369368
 
[1000 rows x 5 columns]

  2.12.2 HDF5

?
1
In [143]: df.to_hdf('foo.h5','df')

  读入

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [144]: pd.read_hdf('foo.h5','df')
Out[144]:
                    A          B         C          D
2000-01-01   0.266457  -0.399641 -0.219582   1.186860
2000-01-02  -1.170732  -0.345873  1.653061  -0.282953
2000-01-03  -1.734933   0.530468  2.060811  -0.515536
2000-01-04  -1.555121   1.452620  0.239859  -1.156896
2000-01-05   0.578117   0.511371  0.103552  -2.428202
2000-01-06   0.478344   0.449933 -0.741620  -1.962409
2000-01-07   1.235339  -0.091757 -1.543861  -1.084753
...               ...        ...       ...        ...
2002-09-20 -10.628548  -9.153563 -7.883146  28.313940
2002-09-21 -10.390377  -8.727491 -6.399645  30.914107
2002-09-22  -8.985362  -8.485624 -4.669462  31.367740
2002-09-23  -9.558560  -8.781216 -4.499815  30.518439
2002-09-24  -9.902058  -9.340490 -4.386639  30.105593
2002-09-25 -10.216020  -9.480682 -3.933802  29.758560
2002-09-26 -11.856774 -10.671012 -3.216025  29.369368
 
[1000 rows x 4 columns]

  2.12.3 Excel

?
1
In [145]: df.to_excel('foo.xlsx', sheet_name='Sheet1')

  读入

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [146]: pd.read_excel('foo.xlsx', 'Sheet1', index_col=None, na_values=['NA'])
Out[146]:
                    A          B         C          D
2000-01-01   0.266457  -0.399641 -0.219582   1.186860
2000-01-02  -1.170732  -0.345873  1.653061  -0.282953
2000-01-03  -1.734933   0.530468  2.060811  -0.515536
2000-01-04  -1.555121   1.452620  0.239859  -1.156896
2000-01-05   0.578117   0.511371  0.103552  -2.428202
2000-01-06   0.478344   0.449933 -0.741620  -1.962409
2000-01-07   1.235339  -0.091757 -1.543861  -1.084753
...               ...        ...       ...        ...
2002-09-20 -10.628548  -9.153563 -7.883146  28.313940
2002-09-21 -10.390377  -8.727491 -6.399645  30.914107
2002-09-22  -8.985362  -8.485624 -4.669462  31.367740
2002-09-23  -9.558560  -8.781216 -4.499815  30.518439
2002-09-24  -9.902058  -9.340490 -4.386639  30.105593
2002-09-25 -10.216020  -9.480682 -3.933802  29.758560
2002-09-26 -11.856774 -10.671012 -3.216025  29.369368
 
[1000 rows x 4 columns]

  

 官方文档:http://pandas.pydata.org/pandas-docs/stable/10min.html#histogramming

 

 

  

 

 

 

 

 

 

  

  

 

这篇文章主要介绍了汇编语言AAD,DAA,DAS用法解析,文中通过示例代码介绍的非常详细,对大家的学习或者工作具有一定的参考学习价值,需要的朋友可以参考下AAD为下面的除法做准备,把AL和AH两个单
摘要:数字化时代下,华为云数据管理服务DAS基于AI技术,于近期推出了智能SQL分析(包括慢SQL发现、SQL透视)、workload级别索引推荐、存储空间预测等运维特性,加上原有的数据库运维能力,
DAShttps://wwwprimaresearchorg/das2016/12thDocumentAnalysisSystemshttps://das2018cvltuwienacat/en/13
一、存储的分类根据服务器类型分为:封闭系统的存储和开放系统的存储,封闭系统主要指大型机,开放系统指基于Windows、UNIX、Linux等操作系统的服务器;开放系统的存储分为:内置存储外挂存储外挂存
PassayearlearninginHangzhou,yifenfeiarrivalhometownNingboatfinallyLeaveNingbooneyear,yifenfeihaveman
C(n,m)从n个字符中选取m个字符,获得所有的组合用编程实现C(n,m)组合问题,可以用递归的方法的解决,将n个字符排列成一条流水线,然后从第一个字符开始选取,并且将已经选取的字符进行保存,如果已经
A(M,N)是从M个不同元素中选N个元素的排列数(区分元素排列先后顺序)C(M,N)是从M个不同元素中选N个元素的组合数(不区分元素排列先后顺序)C(n,m)n!/m!(nm)!故C(4,1)4!/3
摘要:华为云数据管理服务DAS提供的云DBA智能化运维平台于日前发布了空间元数据分析特性,支持查看实例空间异常列表、实例空间概况、数据库列表信息,帮助用户及时发现数据库中的空间和元数据异常,有效保障
来源:学生作业帮编辑:作业帮分类:综合作业时间:2021/05/1821:45:03C语言题目:输入两个矩阵,a[m][n]和b[n][p];求他们的乘积描述:输入两个个矩阵的行列mn;并输入其数值如
易语言求N次方,希望大家都能学懂。非无私奉献,拒绝看经验不回复,学懂了记得按照下面做:在底部点击相关按钮,给我投票和点赞。有心得或者疑问,点击相关按钮反馈。具体操作请看图。1、易语言新建一个wind