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
- 通过标签选取
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)