九色国产,午夜在线视频,新黄色网址,九九色综合,天天做夜夜做久久做狠狠,天天躁夜夜躁狠狠躁2021a,久久不卡一区二区三区

打開APP
userphoto
未登錄

開通VIP，暢享免費(fèi)電子書等14項(xiàng)超值服

開通VIP

首頁

好書

留言交流

下載APP

聯(lián)系客服
十分鐘搞定pandas
本文來源（伯樂在線）：http://python.jobbole.com/84416/

原文出處： pandas.pydata.org   譯文出處：石卓林   

這是關(guān)于pandas的簡短介紹，主要面向新用戶。可以參閱Cookbook了解更復(fù)雜的使用方法。

習(xí)慣上,我們做以下導(dǎo)入

Python
1
2
3
In [1]: import pandas as pd
In [2]: import numpy as np
In [3]: import matplotlib.pyplot as plt

創(chuàng)建對象

使用傳遞的值列表序列創(chuàng)建序列, 讓pandas創(chuàng)建默認(rèn)整數(shù)索引

Python
1
2
3
4
5
6
7
8
9
10
In [4]: s = pd.Series([1,3,5,np.nan,6,8])
In [5]: s
Out[5]:
0     1
1     3
2     5
3   NaN
4     6
5     8
dtype: float64

使用傳遞的numpy數(shù)組創(chuàng)建數(shù)據(jù)幀,并使用日期索引和標(biāo)記列.

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
In [6]: dates = pd.date_range('20130101',periods=6)
In [7]: dates
Out[7]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01, ..., 2013-01-06]
Length: 6, Freq: D, Timezone: None
In [8]: df = pd.DataFrame(np.random.randn(6,4),index=dates,columns=list('ABCD'))
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

使用傳遞的可轉(zhuǎn)換序列的字典對象創(chuàng)建數(shù)據(jù)幀.

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
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 2013-01-02  1  3   test  foo
1  1 2013-01-02  1  3  train  foo
2  1 2013-01-02  1  3   test  foo
3  1 2013-01-02  1  3  train  foo

所有明確類型

Python
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

如果你這個正在使用IPython，標(biāo)簽補(bǔ)全列名（以及公共屬性）將自動啟用。這里是將要完成的屬性的子集：

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
In [13]: df2.<TAB>
df2.A                  df2.boxplot
df2.abs                df2.C
df2.add                df2.clip
df2.add_prefix         df2.clip_lower
df2.add_suffix         df2.clip_upper
df2.align              df2.columns
df2.all                df2.combine
df2.any                df2.combineAdd
df2.append             df2.combine_first
df2.apply              df2.combineMult
df2.applymap           df2.compound
df2.as_blocks          df2.consolidate
df2.asfreq             df2.convert_objects
df2.as_matrix          df2.copy
df2.astype             df2.corr
df2.at                 df2.corrwith
df2.at_time            df2.count
df2.axes               df2.cov
df2.B                  df2.cummax
df2.between_time       df2.cummin
df2.bfill              df2.cumprod
df2.blocks             df2.cumsum
df2.bool               df2.D

如你所見, 列 A, B, C, 和 D 也是自動完成標(biāo)簽. E 也是可用的; 為了簡便起見,后面的屬性顯示被截?cái)?

查看數(shù)據(jù)

參閱基礎(chǔ)部分

查看幀頂部和底部行

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
In [14]: df.head()
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)
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

顯示索引,列,和底層numpy數(shù)據(jù)

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
In [16]: df.index
Out[16]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01, ..., 2013-01-06]
Length: 6, Freq: D, Timezone: None
In [17]: df.columns
Out[17]: Index([u'A', u'B', u'C', u'D'], dtype='object')
In [18]: df.values
Out[18]:
array([[ 0.4691, -0.2829, -1.5091, -1.1356],
       [ 1.2121, -0.1732,  0.1192, -1.0442],
       [-0.8618, -2.1046, -0.4949,  1.0718],
       [ 0.7216, -0.7068, -1.0396,  0.2719],
       [-0.425 ,  0.567 ,  0.2762, -1.0874],
       [-0.6737,  0.1136, -1.4784,  0.525 ]])

描述顯示數(shù)據(jù)快速統(tǒng)計(jì)摘要

Python
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

轉(zhuǎn)置數(shù)據(jù)

Python
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

按軸排序

Python
1
2
3
4
5
6
7
8
9
In [21]: df.sort_index(axis=1, ascending=False)
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

Python
1
2
3
4
5
6
7
8
9
In [22]: df.sort(columns='B')
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

選擇器

注釋: 標(biāo)準(zhǔn)Python / Numpy表達(dá)式可以完成這些互動工作, 但在生產(chǎn)代碼中, 我們推薦使用優(yōu)化的pandas數(shù)據(jù)訪問方法, .at, .iat, .loc, .iloc 和 .ix.

參閱索引文檔 索引和選擇數(shù)據(jù) and 多索引/高級索引

讀取

選擇單列, 這會產(chǎn)生一個序列, 等價df.A

Python
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

使用[]選擇行片斷

Python
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

使用標(biāo)簽選擇

更多信息請參閱按標(biāo)簽選擇

使用標(biāo)簽獲取橫截面

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

使用標(biāo)簽選擇多軸

Python
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

顯示標(biāo)簽切片, 包含兩個端點(diǎn)

Python
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

降低返回對象維度

Python
1
2
3
4
5
In [29]: df.loc['20130102',['A','B']]
Out[29]:
A    1.212112
B   -0.173215
Name: 2013-01-02 00:00:00, dtype: float64

獲取標(biāo)量值

Python
1
2
In [30]: df.loc[dates[0],'A']
Out[30]: 0.46911229990718628

快速訪問并獲取標(biāo)量數(shù)據(jù) (等價上面的方法)

Python
1
2
In [31]: df.at[dates[0],'A']
Out[31]: 0.46911229990718628

按位置選擇

更多信息請參閱按位置參閱

傳遞整數(shù)選擇位置

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

使用整數(shù)片斷,效果類似numpy/python

Python
1
2
3
4
5
In [33]: df.iloc[3:5,0:2]
Out[33]:
                   A         B
2013-01-04  0.721555 -0.706771
2013-01-05 -0.424972  0.567020

使用整數(shù)偏移定位列表,效果類似 numpy/python 樣式

Python
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

顯式行切片

Python
1
2
3
4
5
In [35]: df.iloc[1:3,:]
Out[35]:
                   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

顯式列切片

Python
1
2
3
4
5
6
7
8
9
In [36]: df.iloc[:,1:3]
Out[36]:
                   B         C
2013-01-01 -0.282863 -1.509059
2013-01-02 -0.173215  0.119209
2013-01-03 -2.104569 -0.494929
2013-01-04 -0.706771 -1.039575
2013-01-05  0.567020  0.276232
2013-01-06  0.113648 -1.478427

顯式獲取一個值

Python
1
2
In [37]: df.iloc[1,1]
Out[37]: -0.17321464905330861

快速訪問一個標(biāo)量（等同上個方法）

Python
1
2
In [38]: df.iat[1,1]
Out[38]: -0.17321464905330861

布爾索引

使用單個列的值選擇數(shù)據(jù).

Python
1
2
3
4
5
6
In [39]: df[df.A > 0]
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

where 操作.

Python
1
2
3
4
5
6
7
8
9
In [40]: df[df > 0]
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

使用 isin() 篩選：

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
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

賦值

賦值一個新列，通過索引自動對齊數(shù)據(jù)

Python
1
2
3
4
5
6
7
8
9
10
11
12
In [45]: s1 = pd.Series([1,2,3,4,5,6],index=pd.date_range('20130102',periods=6))
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

按標(biāo)簽賦值

Python
1
In [48]: df.at[dates[0],'A'] = 0

按位置賦值

Python
1
In [49]: df.iat[0,1] = 0

通過numpy數(shù)組分配賦值

Python
1
In [50]: df.loc[:,'D'] = np.array([5] * len(df))

之前的操作結(jié)果

Python
1
2
3
4
5
6
7
8
9
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
2013-01-03 -0.861849 -2.104569 -0.494929  5   2
2013-01-04  0.721555 -0.706771 -1.039575  5   3
2013-01-05 -0.424972  0.567020  0.276232  5   4
2013-01-06 -0.673690  0.113648 -1.478427  5   5

where 操作賦值.

Python
1
2
3
4
5
6
7
8
9
10
11
In [52]: df2 = df.copy()
In [53]: df2[df2 > 0] = -df2
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
2013-01-03 -0.861849 -2.104569 -0.494929 -5  -2
2013-01-04 -0.721555 -0.706771 -1.039575 -5  -3
2013-01-05 -0.424972 -0.567020 -0.276232 -5  -4
2013-01-06 -0.673690 -0.113648 -1.478427 -5  -5

丟失的數(shù)據(jù)

pandas主要使用np.nan替換丟失的數(shù)據(jù). 默認(rèn)情況下它并不包含在計(jì)算中. 請參閱 Missing Data section

重建索引允許更改/添加/刪除指定軸索引,并返回?cái)?shù)據(jù)副本.

Python
1
2
3
4
5
6
7
8
9
In [55]: df1 = df.reindex(index=dates[0:4],columns=list(df.columns) + ['E'])
In [56]: df1.loc[dates[0]:dates[1],'E'] = 1
In [57]: df1
Out[57]:
                   A         B         C  D   F   E
2013-01-01  0.000000  0.000000 -1.509059  5 NaN   1
2013-01-02  1.212112 -0.173215  0.119209  5   1   1
2013-01-03 -0.861849 -2.104569 -0.494929  5   2 NaN
2013-01-04  0.721555 -0.706771 -1.039575  5   3 NaN

刪除任何有丟失數(shù)據(jù)的行.

Python
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  1

填充丟失數(shù)據(jù)

Python
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  1
2013-01-02  1.212112 -0.173215  0.119209  5  1  1
2013-01-03 -0.861849 -2.104569 -0.494929  5  2  5
2013-01-04  0.721555 -0.706771 -1.039575  5  3  5

獲取值是否nan的布爾標(biāo)記

Python
1
2
3
4
5
6
7
In [60]: pd.isnull(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

運(yùn)算

參閱二元運(yùn)算基礎(chǔ)

統(tǒng)計(jì)

計(jì)算時一般不包括丟失的數(shù)據(jù)

執(zhí)行描述性統(tǒng)計(jì)

Python
1
2
3
4
5
6
7
8
In [61]: df.mean()
Out[61]:
A   -0.004474
B   -0.383981
C   -0.687758
D    5.000000
F    3.000000
dtype: float64

在其他軸做相同的運(yùn)算

Python
1
2
3
4
5
6
7
8
9
In [62]: df.mean(1)
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

用于運(yùn)算的對象有不同的維度并需要對齊.除此之外，pandas會自動沿著指定維度計(jì)算.

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [63]: s = pd.Series([1,3,5,np.nan,6,8],index=dates).shift(2)
In [64]: s
Out[64]:
2013-01-01   NaN
2013-01-02   NaN
2013-01-03     1
2013-01-04     3
2013-01-05     5
2013-01-06   NaN
Freq: D, dtype: float64
In [65]: df.sub(s,axis='index')
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   1
2013-01-04 -2.278445 -3.706771 -4.039575   2   0
2013-01-05 -5.424972 -4.432980 -4.723768   0  -1
2013-01-06       NaN       NaN       NaN NaN NaN

Apply

在數(shù)據(jù)上使用函數(shù)

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
In [66]: df.apply(np.cumsum)
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
2013-01-03  0.350263 -2.277784 -1.884779  15   3
2013-01-04  1.071818 -2.984555 -2.924354  20   6
2013-01-05  0.646846 -2.417535 -2.648122  25  10
2013-01-06 -0.026844 -2.303886 -4.126549  30  15
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

直方圖

請參閱 直方圖和離散化

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
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: int32
In [70]: s.value_counts()
Out[70]:
4    5
6    2
2    2
1    1
dtype: int64

字符串方法

序列可以使用一些字符串處理方法很輕易操作數(shù)據(jù)組中的每個元素,比如以下代碼片斷。 注意字符匹配方法默認(rèn)情況下通常使用正則表達(dá)式（并且大多數(shù)時候都如此）. 更多信息請參閱字符串向量方法.

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

合并

連接

pandas提供各種工具以簡便合并序列,數(shù)據(jù)楨,和組合對象, 在連接/合并類型操作中使用多種類型索引和相關(guān)數(shù)學(xué)函數(shù).

請參閱合并部分

把pandas對象連接到一起

Python
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
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 [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

連接

SQL樣式合并. 請參閱 數(shù)據(jù)庫style聯(lián)接

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
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

添加

添加行到數(shù)據(jù)增. 參閱 添加

Python
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 [82]: df = pd.DataFrame(np.random.randn(8, 4), columns=['A','B','C','D'])
In [83]: df
Out[83]:
          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 [84]: s = df.iloc[3]
In [85]: df.append(s, ignore_index=True)
Out[85]:
          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

分組

對于“group by”指的是以下一個或多個處理

  • 將數(shù)據(jù)按某些標(biāo)準(zhǔn)分割為不同的組
  • 在每個獨(dú)立組上應(yīng)用函數(shù)
  • 組合結(jié)果為一個數(shù)據(jù)結(jié)構(gòu)

請參閱 分組部分

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
In [86]: 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 [87]: df
Out[87]:
     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

分組然后應(yīng)用函數(shù)統(tǒng)計(jì)總和存放到結(jié)果組

Python
1
2
3
4
5
6
In [88]: df.groupby('A').sum()
Out[88]:
            C        D
A                    
bar -2.802588  2.42611
foo  3.146492 -0.63958

按多列分組為層次索引,然后應(yīng)用函數(shù)

Python
1
2
3
4
5
6
7
8
9
10
In [89]: df.groupby(['A','B']).sum()
Out[89]:
                  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

重塑

請參閱章節(jié) 分層索引 和 重塑.

堆疊

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
In [90]: tuples = list(zip(*[['bar', 'bar', 'baz', 'baz',
   ....:                      'foo', 'foo', 'qux', 'qux'],
   ....:                     ['one', 'two', 'one', 'two',
   ....:                      'one', 'two', 'one', 'two']]))
   ....:
In [91]: index = pd.MultiIndex.from_tuples(tuples, names=['first', 'second'])
In [92]: df = pd.DataFrame(np.random.randn(8, 2), index=index, columns=['A', 'B'])
In [93]: df2 = df[:4]
In [94]: df2
Out[94]:
                     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

堆疊 函數(shù) “壓縮” 數(shù)據(jù)楨的列一個級別.

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
In [95]: stacked = df2.stack()
In [96]: stacked
Out[96]:
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

被“堆疊”數(shù)據(jù)楨或序列(有多個索引作為索引), 其堆疊的反向操作是未堆棧, 上面的數(shù)據(jù)默認(rèn)反堆疊到上一級別:

Python
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 [97]: stacked.unstack()
Out[97]:
                     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 [98]: stacked.unstack(1)
Out[98]:
second        one       two
first                      
bar   A  0.029399  0.282696
      B -0.542108 -0.087302
baz   A -1.575170  0.816482
      B  1.771208  1.100230
In [99]: stacked.unstack(0)
Out[99]:
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

數(shù)據(jù)透視表

查看數(shù)據(jù)透視表.

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
In [100]: 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 [101]: df
Out[101]:
        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

我們可以從此數(shù)據(jù)非常容易的產(chǎn)生數(shù)據(jù)透視表:

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
In [102]: pd.pivot_table(df, values='D', index=['A', 'B'], columns=['C'])
Out[102]:
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
      C  0.648740       NaN
two   A       NaN  0.100900
      B -1.170653       NaN
      C       NaN  0.536826

時間序列

pandas有易用,強(qiáng)大且高效的函數(shù)用于高頻數(shù)據(jù)重采樣轉(zhuǎn)換操作(例如,轉(zhuǎn)換秒數(shù)據(jù)到5分鐘數(shù)據(jù)), 這是很普遍的情況，但并不局限于金融應(yīng)用, 請參閱時間序列章節(jié)

Python
1
2
3
4
5
6
In [103]: rng = pd.date_range('1/1/2012', periods=100, freq='S')
In [104]: ts = pd.Series(np.random.randint(0, 500, len(rng)), index=rng)
In [105]: ts.resample('5Min', how='sum')
Out[105]:
2012-01-01    25083
Freq: 5T, dtype: int32

時區(qū)表示

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [106]: rng = pd.date_range('3/6/2012 00:00', periods=5, freq='D')
In [107]: ts = pd.Series(np.random.randn(len(rng)), rng)
In [108]: ts
Out[108]:
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 [109]: ts_utc = ts.tz_localize('UTC')
In [110]: ts_utc
Out[110]:
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

轉(zhuǎn)換到其它時區(qū)

Python
1
2
3
4
5
6
7
8
In [111]: ts_utc.tz_convert('US/Eastern')
Out[111]:
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

轉(zhuǎn)換不同的時間跨度

Python
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
In [112]: rng = pd.date_range('1/1/2012', periods=5, freq='M')
In [113]: ts = pd.Series(np.random.randn(len(rng)), index=rng)
In [114]: ts
Out[114]:
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 [115]: ps = ts.to_period()
In [116]: ps
Out[116]:
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 [117]: ps.to_timestamp()
Out[117]:
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

轉(zhuǎn)換時段并且使用一些運(yùn)算函數(shù), 下例中, 我們轉(zhuǎn)換年報(bào)11月到季度結(jié)束每日上午9點(diǎn)數(shù)據(jù)

Python
1
2
3
4
5
6
7
8
9
10
11
In [118]: prng = pd.period_range('1990Q1', '2000Q4', freq='Q-NOV')
In [119]: ts = pd.Series(np.random.randn(len(prng)), prng)
In [120]: ts.index = (prng.asfreq('M', 'e') + 1).asfreq('H', 's') + 9
In [121]: ts.head()
Out[121]:
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

分類

自版本0.15起, pandas可以在數(shù)據(jù)楨中包含分類. 完整的文檔, 請查看分類介紹 and the API文檔.

Python
1
In [122]: df = pd.DataFrame({"id":[1,2,3,4,5,6], "raw_grade":['a', 'b', 'b', 'a', 'a', 'e']})

轉(zhuǎn)換原始類別為分類數(shù)據(jù)類型.

Python
1
2
3
4
5
6
7
8
9
10
11
In [123]: df["grade"] = df["raw_grade"].astype("category")
In [124]: df["grade"]
Out[124]:
0    a
1    b
2    b
3    a
4    a
5    e
Name: grade, dtype: category
Categories (3, object): [a, b, e]

重命令分類為更有意義的名稱 (分配到Series.cat.categories對應(yīng)位置!)

Python
1
In [125]: df["grade"].cat.categories = ["very good", "good", "very bad"]

重排順分類,同時添加缺少的分類(序列 .cat方法下返回新默認(rèn)序列)

Python
1
2
3
4
5
6
7
8
9
10
11
In [126]: df["grade"] = df["grade"].cat.set_categories(["very bad", "bad", "medium", "good", "very good"])
In [127]: df["grade"]
Out[127]:
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]

排列分類中的順序,不是按詞匯排列.

Python
1
2
3
4
5
6
7
8
9
In [128]: df.sort("grade")
Out[128]:
   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

類別列分組,并且也顯示空類別.

Python
1
2
3
4
5
6
7
8
9
In [129]: df.groupby("grade").size()
Out[129]:
grade
very bad      1
bad         NaN
medium      NaN
good          2
very good     3
dtype: float64

繪圖

繪圖文檔.

Python
1
2
3
4
In [130]: ts = pd.Series(np.random.randn(1000), index=pd.date_range('1/1/2000', periods=1000))
In [131]: ts = ts.cumsum()
In [132]: ts.plot()
Out[132]: <matplotlib.axes._subplots.AxesSubplot at 0xb02091ac>

在數(shù)據(jù)楨中,可以很方便的繪制帶標(biāo)簽列:

Python
1
2
3
4
5
6
In [133]: df = pd.DataFrame(np.random.randn(1000, 4), index=ts.index,
   .....:                   columns=['A', 'B', 'C', 'D'])
   .....:
In [134]: df = df.cumsum()
In [135]: plt.figure(); df.plot(); plt.legend(loc='best')
Out[135]: <matplotlib.legend.Legend at 0xb01c9cac>

獲取數(shù)據(jù)輸入/輸出

CSV

寫入csv文件

Python
1
In [136]: df.to_csv('foo.csv')

讀取csv文件

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [137]: pd.read_csv('foo.csv')
Out[137]:
     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]

HDF5

讀寫HDF存儲

寫入HDF5存儲

Python
1
In [138]: df.to_hdf('foo.h5','df')

讀取HDF5存儲

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [139]: pd.read_hdf('foo.h5','df')
Out[139]:
                    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]

Excel

讀寫MS Excel

寫入excel文件

Python
1
In [140]: df.to_excel('foo.xlsx', sheet_name='Sheet1')

讀取excel文件

Python
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
In [141]: pd.read_excel('foo.xlsx', 'Sheet1', index_col=None, na_values=['NA'])
Out[141]:
                    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]

陷阱

如果嘗試這樣操作可能會看到像這樣的異常:

Python
1
2
3
4
5
>>> if pd.Series([False, True, False]):
    print("I was true")
Traceback
    ...
ValueError: The truth value of an array is ambiguous. Use a.empty, a.any() or a.all().

查看對照獲取解釋和怎么做的幫助

也可以查看陷阱.

2 贊 10 收藏 評論
本站僅提供存儲服務(wù)，所有內(nèi)容均由用戶發(fā)布，如發(fā)現(xiàn)有害或侵權(quán)內(nèi)容，請點(diǎn)擊舉報(bào)。
打開APP，閱讀全文并永久保存 查看更多類似文章
猜你喜歡
類似文章
pandas 基礎(chǔ)操作 更新
PANDAS QUICK START 
[轉(zhuǎn)]10 minutes to pandas
<font style="vertical-align: inherit;"><font style="vertical-align: inherit;
簡約而不簡單|值得收藏的Pandas基本操作指南
python小技能-缺失數(shù)據(jù)處理
更多類似文章 >>
生活服務(wù)
熱點(diǎn)新聞
分享 收藏 導(dǎo)長圖 關(guān)注 下載文章
綁定賬號成功
后續(xù)可登錄賬號暢享VIP特權(quán)！
如果VIP功能使用有故障，
可點(diǎn)擊這里聯(lián)系客服！

聯(lián)系客服