pandas 教程

Pandas 数据结构

Pandas 基本操作

Pandas API

Pandas快速入门


这是一个Pandas快速入门教程,主要面向新用户。这里主要是为那些喜欢“短平快”的读者准备的,有兴趣的读者可通过其它教程文章来一步一步地更复杂的应用知识。

首先,假设您安装好了Anaconda,现在启动Anaconda开始学始本教程中的示例。工作界面如下所示 -

测试工作环境是否有安装好了Pandas,导入相关包如下:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
print("Hello, Pandas")

然后执行一下,看有没有问题,如果正常应该会在终端输出区看到以下结果 -

对象创建

通过传递值列表来创建一个系列,让Pandas创建一个默认的整数索引:

import pandas as pd
import numpy as np

s = pd.Series([1,3,5,np.nan,6,8])

print(s)

执行后输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
0    1.0
1    3.0
2    5.0
3    NaN
4    6.0
5    8.0
dtype: float64

通过传递numpy数组,使用datetime索引和标记列来创建DataFrame

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=7)
print(dates)

print("--"*16)
df = pd.DataFrame(np.random.randn(7,4), index=dates, columns=list('ABCD'))
print(df)

执行后输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
DatetimeIndex(['2017-01-01', '2017-01-02', '2017-01-03', '2017-01-04',
               '2017-01-05', '2017-01-06', '2017-01-07'],
              dtype='datetime64[ns]', freq='D')
--------------------------------
                   A         B         C         D
2017-01-01 -0.732038  0.329773 -0.156383  0.270800
2017-01-02  0.750144  0.722037 -0.849848 -1.105319
2017-01-03 -0.786664 -0.204211  1.246395  0.292975
2017-01-04 -1.108991  2.228375  0.079700 -1.738507
2017-01-05  0.348526 -0.960212  0.190978 -2.223966
2017-01-06 -0.579689 -1.355910  0.095982  1.233833
2017-01-07  1.086872  0.664982  0.377787  1.012772

通过传递可以转换为类似系列的对象的字典来创建 DataFrame。参考以下示例代码 -

import pandas as pd
import numpy as np

df2 = pd.DataFrame({ 'A' : 1.,
                     'B' : pd.Timestamp('20170102'),
                     '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' })

print(df2)

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
     A          B    C  D      E    F
0  1.0 2017-01-02  1.0  3   test  foo
1  1.0 2017-01-02  1.0  3  train  foo
2  1.0 2017-01-02  1.0  3   test  foo
3  1.0 2017-01-02  1.0  3  train  foo

有指定dtypes,参考以下示例代码 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
A           float64
B    datetime64[ns]
C           float32
D             int32
E          category
F            object
dtype: object

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
A           float64
B    datetime64[ns]
C           float32
D             int32
E          category
F            object
dtype: object

如果使用 IPython,则会自动启用列名(以及公共属性)的选项完成。以下是将要完成的属性的一个子集:

In [13]: df2.<TAB>
df2.A                  df2.bool
df2.abs                df2.boxplot
df2.add                df2.C
df2.add_prefix         df2.clip
df2.add_suffix         df2.clip_lower
df2.align              df2.clip_upper
df2.all                df2.columns
df2.any                df2.combine
df2.append             df2.combine_first
df2.apply              df2.compound
df2.applymap           df2.consolidate
df2.D

可以看到,列ABCD自动标签完成。E也在一样。其余的属性为了简洁而被截短。

查看数据

查看框架的顶部和底部的数据行。参考以下示例代码 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=7)
df = pd.DataFrame(np.random.randn(7,4), index=dates, columns=list('ABCD'))
print(df.head())
print("--------------" * 10)
print(df.tail(3))

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
                   A         B         C         D
2017-01-01 -0.520856 -0.555019 -2.286424  1.745681
2017-01-02  1.114030  0.861933  0.795958  0.420670
2017-01-03 -0.343605 -0.937356  0.052693 -0.540735
2017-01-04  1.587684 -0.743756  0.021738 -0.702190
2017-01-05  1.243403  0.930299  0.234343  1.604182
------------------------------------------------------------
                   A         B         C         D
2017-01-05  1.243403  0.930299  0.234343  1.604182
2017-01-06 -0.087004 -0.368055  1.434022  0.464193
2017-01-07 -1.248981  0.973724 -0.288384 -0.577388

显示索引,列和底层 numpy 数据,参考以下代码 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=7)
df = pd.DataFrame(np.random.randn(7,4), index=dates, columns=list('ABCD'))
print("index is :" )
print(df.index)
print("columns is :" )
print(df.columns)
print("values is :" )
print(df.values)

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
index is :
DatetimeIndex(['2017-01-01', '2017-01-02', '2017-01-03', '2017-01-04',
               '2017-01-05', '2017-01-06', '2017-01-07'],
              dtype='datetime64[ns]', freq='D')
columns is :
Index(['A', 'B', 'C', 'D'], dtype='object')
values is :
[[ 2.23820398  0.18440123  0.08039084 -0.27751926]
 [-0.12335513  0.36641304 -0.28617579  0.34383109]
 [-0.85403491  0.63876989  1.26032173 -1.27382333]
 [-0.07262661 -0.01788962  0.28748668  1.12715561]
 [-1.14293392 -0.88263364  0.72250762 -1.64051326]
 [ 0.41864083  0.40545953 -0.14591541 -0.57168728]
 [ 1.01383531 -0.22793823 -0.44045634  1.04799829]]

描述显示数据的快速统计摘要,参考以下示例代码 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=7)
df = pd.DataFrame(np.random.randn(7,4), index=dates, columns=list('ABCD'))
print(df.describe())

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
              A         B         C         D
count  7.000000  7.000000  7.000000  7.000000
mean  -0.675425 -0.257835  0.144049  0.275621
std    1.697957  0.793953  1.301520  1.412291
min   -2.595040 -1.200401 -1.230538 -0.976166
25%   -1.992393 -0.723464 -0.897041 -0.800919
50%   -1.050666 -0.445612  0.004719 -0.705840
75%    0.592677  0.068574  0.874195  1.398337
max    1.717166  1.150948  2.279856  2.416514

调换数据,参考以下示例代码 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))
print(df.T)

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
   2017-01-01  2017-01-02  2017-01-03  2017-01-04  2017-01-05  2017-01-06
A    0.932454   -2.148503    1.398975    1.565676   -0.167527   -0.242041
B    0.584585    1.373330   -0.069801   -0.102857    1.286432   -0.703491
C   -0.345119   -0.680955    1.686750    1.184996    0.016170   -0.663963
D    0.431751    0.444830   -1.524739    0.040007    0.220172    1.423627

通过轴排序,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))
print(df.sort_index(axis=1, ascending=False))
`

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
                   D         C         B         A
2017-01-01  0.426359  2.542352 -0.324047  0.418973
2017-01-02 -0.834625 -1.356709  0.150744 -1.690500
2017-01-03 -0.018274  0.900801  1.072851  0.149830
2017-01-04 -1.075027 -0.889379 -0.663223 -1.404002
2017-01-05 -1.273966 -1.335761 -1.356561 -1.135199
2017-01-06 -1.590793  0.693430 -0.504164  0.143386

按值排序,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))
print(df.sort_values(by='B'))
`

执行上面示例代码后,输出结果如下 -

                   A         B         C         D
2017-01-06  0.764517 -1.526019  0.400456 -0.182082
2017-01-05 -0.177845 -1.269836 -0.534676  0.796666
2017-01-04 -0.981485 -0.082572 -1.272123  0.508929
2017-01-02 -0.290117  0.053005 -0.295628 -0.346965
2017-01-03  0.941131  0.799280  2.054011 -0.684088
2017-01-01  0.597788  0.892008  0.903053 -0.821024

选择区块

注意虽然用于选择和设置的标准 Python/Numpy 表达式是直观的,可用于交互式工作,但对于生产代码,但建议使用优化的Pandas数据访问方法.at.iat.loc.iloc.ix

获取

选择一列,产生一个系列,相当于df.A,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df['A'])
`

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
2017-01-01    0.317460
2017-01-02   -0.933726
2017-01-03    0.167860
2017-01-04    0.816184
2017-01-05   -0.745503
2017-01-06    0.505319
Freq: D, Name: A, dtype: float64

选择通过[]操作符,选择切片行。参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df[0:3])

print("========= 指定选择日期 ========")

print(df['20170102':'20170103'])
`

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
                   A         B         C         D
2017-01-01  1.103449  0.926571 -1.649978 -0.309270
2017-01-02  0.516404 -0.734076 -0.081163 -0.528497
2017-01-03  0.240356  0.231224 -1.463315 -1.157256
========= 指定选择日期 ========
                   A         B         C         D
2017-01-02  0.516404 -0.734076 -0.081163 -0.528497
2017-01-03  0.240356  0.231224 -1.463315 -1.157256

按标签选择

使用标签获取横截面,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.loc[dates[0]])
`

执行上面示例代码后,输出结果如下 -

runfile('C:/Users/Administrator/.spyder-py3/temp.py', wdir='C:/Users/Administrator/.spyder-py3')
A   -0.483292
B   -0.536987
C   -0.889947
D    1.250857
Name: 2017-01-01 00:00:00, dtype: float64

通过标签选择多轴,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.loc[:,['A','B']])
`

执行上面示例代码后,输出结果如下 -

                   A         B
2017-01-01  0.479048 -0.105106
2017-01-02  0.172920  0.086570
2017-01-03 -1.302485 -0.593550
2017-01-04 -0.595438  1.304054
2017-01-05  0.154267  1.336219
2017-01-06 -0.341204  0.781300

显示标签切片,包括两个端点,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.loc['20170102':'20170104',['A','B']])
`

执行上面示例代码后,输出结果如下 -

                   A         B
2017-01-02  1.062995 -0.108277
2017-01-03  1.962106 -0.294664
2017-01-04 -0.128576  0.717738

减少返回对象的尺寸(大小),参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.loc['20170102',['A','B']])
`

执行上面示例代码后,输出结果如下 -

A    0.252749
B    0.119747
Name: 2017-01-02 00:00:00, dtype: float64

获得标量值,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.loc[dates[0],'A'])
`

执行上面示例代码后,输出结果如下 -

-0.0839903627822

快速访问标量(等同于先前的方法),参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.at[dates[0],'A'])
`

执行上面示例代码后,输出结果如下 -

-0.0839903627822

通过位置选择

通过传递的整数的位置选择,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[3])
`

执行上面示例代码后,输出结果如下 -

A    0.944506
B    1.035781
C    0.421373
D    0.017660
Name: 2017-01-04 00:00:00, dtype: float64

通过整数切片,类似于numpy/python,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[3:5,0:2])
`

执行上面示例代码后,输出结果如下 -

                   A         B
2017-01-04 -1.617068  0.548090
2017-01-05 -0.864247  0.419743

通过整数位置的列表,类似于numpy/python样式,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[[1,2,4],[0,2]])
`

执行上面示例代码后,输出结果如下 -

                   A         C
2017-01-02  0.085091  0.568128
2017-01-03  0.729076 -0.451151
2017-01-05 -1.281975 -0.190119

明确切片行,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[1:3,:])
`

执行上面示例代码后,输出结果如下 -

                   A         B         C         D
2017-01-02 -1.123970 -0.010969 -1.076657 -0.538908
2017-01-03 -0.314408  0.004415 -0.356924  0.337539

明确切片列,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[:,1:3])
`

执行上面示例代码后,输出结果如下 -

                   B         C
2017-01-01  0.323663  1.027599
2017-01-02 -0.176624 -0.959683
2017-01-03  0.689698  0.622540
2017-01-04  1.864511  1.023157
2017-01-05  0.964123  2.062503
2017-01-06 -0.375143  0.231328

要明确获取值,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iloc[1,1])
`

执行上面示例代码后,输出结果如下 -

0.829950900219

要快速访问标量(等同于先前的方法),参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df.iat[1,1])
`

执行上面示例代码后,输出结果如下 -

-0.170996002652

布尔索引

使用单列的值来选择数据,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df[df.A > 0])
`

执行上面示例代码后,输出结果如下 -

                   A         B         C         D
2017-01-03  0.276486 -1.003779  0.721863 -0.558061
2017-01-04  1.177206 -0.464778 -0.116442 -0.385712
2017-01-06  0.846665 -1.398207 -0.145356  0.924342

从满足布尔条件的 DataFrame 中选择值。,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

print(df[df > 0])
`

执行上面示例代码后,输出结果如下 -

                   A         B         C         D
2017-01-01       NaN  1.963213  0.643244  0.945643
2017-01-02  0.364237  0.917368       NaN       NaN
2017-01-03  0.702624       NaN  0.088565       NaN
2017-01-04  1.274313       NaN  2.313910       NaN
2017-01-05  2.586315  0.588273       NaN  1.482597
2017-01-06       NaN  0.405928  0.309201       NaN

使用isin()方法进行过滤,参考以下示例程序 -

import pandas as pd
import numpy as np

dates = pd.date_range('20170101', periods=6)
df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

df2 = df.copy()
df2['E'] = ['one', 'one','two','three','four','three']

print(df2)

print("============= start to filter =============== ")

print(df2[df2['E'].isin(['two','four'])])

`

执行上面示例代码后,输出结果如下 -

                   A         B         C         D      E
2017-01-01  0.723399 -0.369247  0.863941 -1.910875    one
2017-01-02 -0.047573 -0.609780  2.130650 -0.019281    one
2017-01-03 -0.566122 -0.850374 -0.031516  0.362023    two
2017-01-04  0.903819 -0.513673  0.118850 -0.351811  three
2017-01-05 -0.485232 -0.864457  1.396835 -1.696083   four
2017-01-06  0.272145 -0.644449 -1.319063 -0.201354  three
============= start to filter =============== 
                   A         B         C         D     E
2017-01-03 -0.566122 -0.850374 -0.031516  0.362023   two
2017-01-05 -0.485232 -0.864457  1.396835 -1.696083  four