Python数据分析_Pandas01_数据框的创建和选取

主要内容：

• 创建数据表
• 查看数据表
• 数据表索引、选取部分数据
  • 通过标签选取.loc
  • 多重索引选取
  • 位置选取.iloc
  • 布尔索引

---

Object Creation 新建数据

• 用list建series序列

In [73]: s = pd.Series([1,3,5,np.nan,6,8])

In [74]: s
Out[74]:
0    1.0
1    3.0
2    5.0
3    NaN
4    6.0
5    8.0
dtype: float64

• 用numpy array建dataframe

In [75]: dates = pd.date_range('20130101', periods=6)

In [76]: df = pd.DataFrame(np.random.randn(6,4), index=dates, columns=list('ABCD'))

In [77]: df
Out[77]:
                   A         B         C         D
2013-01-01 -0.411674  0.273549  0.629843  1.881497
2013-01-02  1.240512  0.970725  0.033099  1.553420
2013-01-03 -0.544326  0.545738 -1.325810  0.130738
2013-01-04  1.044803 -0.117151  0.874583  2.278227
2013-01-05 -2.194728 -2.536257  0.478644  0.057728
2013-01-06 -1.092031  1.249952  1.598761 -0.153423

#---pd.date_range?---
In [115]: pd.date_range(start='12/31/2011', end='12/31/2013', freq='A')
Out[115]: DatetimeIndex(['2011-12-31', '2012-12-31', '2013-12-31'], dtype='datetime64[ns]', freq='A-DEC')

• 用dictionary

In [78]: 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' })
    ...: df2
    ...:
Out[78]:
     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

In [80]: df2.dtypes
Out[80]:
A           float64
B    datetime64[ns]
C           float32
D             int32
E          category
F            object
dtype: object

在ipython中可以使用<tab>键进行自动补充，它会列出数据对象可以执行的操作。

查看数据

df.head()
df.tail(3)
df.index
df.columns              #返回一个这样的东西：pandas.indexes.numeric.Int64Index
df.values               #提取出数据框的数值，返回一个array

---

数据选取

建议 使用pandas的数据选取方法：.at, .iat, .loc, .iloc, .ix. 这些更高效。

df['A']       # 选取某一列，返回一个Series，== df.A，【只能选某一列，不能用":"多选。】

df[0:3]       # 选行
df['20130102':'20130104']

• 通过标签label选取，.loc

  用.loc[]选取数据时，方括号里对应的是：[行,列]（逗号分隔），如果只有一个值，默认是行。可以用“:”。

  In [82]: df
  Out[82]:
                     A         B         C         D
  2013-01-01 -0.411674  0.273549  0.629843  1.881497
  2013-01-02  1.240512  0.970725  0.033099  1.553420
  2013-01-03 -0.544326  0.545738 -1.325810  0.130738
  2013-01-04  1.044803 -0.117151  0.874583  2.278227
  2013-01-05 -2.194728 -2.536257  0.478644  0.057728
  2013-01-06 -1.092031  1.249952  1.598761 -0.153423
  
  In [83]: df.loc[dates[0]]   # 作为index的日期列叫dates
  Out[83]:
  A   -0.411674
  B    0.273549
  C    0.629843
  D    1.881497
  Name: 2013-01-01 00:00:00, dtype: float64
  
  #---对多个维度轴axis进行选取---
  In [84]: df.loc['20130102':'20130104',['A','B']]
  Out[84]:
                     A         B
  2013-01-02  1.240512  0.970725
  2013-01-03 -0.544326  0.545738
  2013-01-04  1.044803 -0.117151
  
  #---选取某个数值---
  In [85]: df.loc[dates[0],'A']
  Out[85]: -0.41167416696608039
  
  In [86]: df.at[dates[0],'A']     # 更高效的做法
  Out[86]: -0.41167416696608039
  
  

• 多重索引的选取

  index有多个维度

  #这里有一个多重索引
  MultiIndex(levels=[[1, 2, 3], ['count', 'mean', 'std', 'min', '5%', '10%', '15.0%', '20%', '25%',
                              '30.0%', '35%', '40%', '45%', '50%', '55.0%', '60.0%', '65%', '70%',
                              '75%', '80%', '85.0%', '90%', '95%', 'max']],
             labels=[[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
                      1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
                      2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2],
                      [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20,
                       21, 22, 23, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
                       18, 19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14,
                       15, 16, 17, 18, 19, 20, 21, 22, 23]],
             names=['label_1', None])
  
  df[columnName]                      #选某一列，或多列（":",[,,,]）
  df.loc[:,columnName]                #选某一列，或多列（":",[,,,]）
  df.loc[1,columnName]             #可以直接用最外层的索引
  df.loc[(1,'std'),columnName]     #多层索引要用tuple。选多行用":"连接tuple。
  df.loc[[(1,'std'),(2,"count")],'feature_001']
  

• 用位置选取：.iloc

  .lic[]，位置索引，方括号里是整数值。同样的用“,”隔开行列。

  In [93]: df.iloc[3]
  Out[93]:
  A    1.044803
  B   -0.117151
  C    0.874583
  D    2.278227
  Name: 2013-01-04 00:00:00, dtype: float64
  
  In [94]: df.iloc[3:5,0:2]
  Out[94]:
                     A         B
  2013-01-04  1.044803 -0.117151
  2013-01-05 -2.194728 -2.536257
  
  In [95]: df.iat[1,1]
  Out[95]: 0.97072539301549565
  

• **布尔索引 **Boolean Indexing

  某一列大于0的数据

  In [96]: df[df.A > 0]
  Out[96]:
                     A         B         C         D
  2013-01-02  1.240512  0.970725  0.033099  1.553420
  2013-01-04  1.044803 -0.117151  0.874583  2.278227
  

  整体大于零的数据。小于0的为NaN

  In [97]: df[df > 0]
  Out[97]:
                     A         B         C         D
  2013-01-01       NaN  0.273549  0.629843  1.881497
  2013-01-02  1.240512  0.970725  0.033099  1.553420
  2013-01-03       NaN  0.545738       NaN  0.130738
  2013-01-04  1.044803       NaN  0.874583  2.278227
  2013-01-05       NaN       NaN  0.478644  0.057728
  2013-01-06       NaN  1.249952  1.598761       NaN
  

  对字符型数据选取

  #---isin ---
  In [98]: df2 = df.copy()
      ...: df2['E'] = ['one', 'one','two','three','four','three']
      ...: df2
      ...:
  Out[98]:
                     A         B         C         D      E
  2013-01-01 -0.411674  0.273549  0.629843  1.881497    one
  2013-01-02  1.240512  0.970725  0.033099  1.553420    one
  2013-01-03 -0.544326  0.545738 -1.325810  0.130738    two
  2013-01-04  1.044803 -0.117151  0.874583  2.278227  three
  2013-01-05 -2.194728 -2.536257  0.478644  0.057728   four
  2013-01-06 -1.092031  1.249952  1.598761 -0.153423  three
  
  In [99]: df2[df2['E'].isin(['two','four'])]
  Out[99]:
                     A         B         C         D     E
  2013-01-03 -0.544326  0.545738 -1.325810  0.130738   two
  2013-01-05 -2.194728 -2.536257  0.478644  0.057728  four
  

  使用布尔面具

  In [107]: mask = df2["A"] >0
  
  In [108]: df3 = df2[mask]
  
  In [109]: df3
  Out[109]:
                     A         B         C         D      E
  2013-01-02  1.240512  0.970725  0.033099  1.553420    ONE
  2013-01-04  1.044803 -0.117151  0.874583  2.278227  THREE
  
  # 查看无重复的值：.unique()
  In [101]: df2.loc[:,"E"].unique()
  Out[101]: array(['one', 'two', 'three', 'four'], dtype=object)