Pandas

一、多重索引

tuples = list(zip(*[['bar', 'bar', 'baz', 'baz',
                     'foo', 'foo', 'qux', 'qux'],
                    ['one', 'two', 'one', 'two',
                    'one', 'two', 'one', 'two']]))
index = pd.MultiIndex.from_tuples(tuples, names=['first', 'second'])
df = pd.DataFrame(np.random.randn(8, 2), index=index, columns=['A', 'B']) #将列索引变成行索引
stacked = df.stack()
print(stacked.index)

MultiIndex([('bar', 'one', 'A'),
            ('bar', 'one', 'B'),
            ('bar', 'two', 'A'),
            ('bar', 'two', 'B'),
            ('baz', 'one', 'A'),
            ('baz', 'one', 'B'),
            ('baz', 'two', 'A'),
            ('baz', 'two', 'B'),
            ('foo', 'one', 'A'),
            ('foo', 'one', 'B'),
            ('foo', 'two', 'A'),
            ('foo', 'two', 'B'),
            ('qux', 'one', 'A'),
            ('qux', 'one', 'B'),
            ('qux', 'two', 'A'),
            ('qux', 'two', 'B')],
           names=['first', 'second', None])

二、数据透视
只看一部分数据

df = pd.DataFrame({'A': ['one', 'one', 'two', 'thee']*3,
                   'B': ['A', 'B', 'C']*4,
                   'C': ['foo', 'foo', 'foo', 'bar', 'bar', 'bar']*2,
                   'D': np.random.randn(12),
                   'E': np.random.randn(12)
                   })
print(df)

      A  B    C         D         E
0    one  A  foo  1.383200 -0.146617
1    one  B  foo  0.055231 -0.654110
2    two  C  foo -0.341115  0.623954
3   thee  A  bar  2.878976  0.833007
4    one  B  bar -0.690180  1.630808
5    one  C  bar  0.686234  0.954150
6    two  A  foo  0.304506  1.406060
7   thee  B  foo -0.105632  0.320874
8    one  C  foo -0.915833  0.175327
9    one  A  bar -0.392937 -0.310135
10   two  B  bar -0.714334  0.831486
11  thee  C  bar  0.808777 -2.476674

#看以AB为行索引，C为列索引，针对E的数据
print(df.pivot_table(values=['D'], index=['A', 'B'], columns=['C']))

现在A、B变成行索引，先通过A的one，A，在从C中找到bar，就可以对应到df中的-0.932937

            D          
C            bar       foo
A    B                    
one  A -0.392937  1.383200
     B -0.690180  0.055231
     C  0.686234 -0.915833
thee A  2.878976       NaN
     B       NaN -0.105632
     C  0.808777       NaN
two  A       NaN  0.304506
     B -0.714334       NaN
     C       NaN -0.341115

Process finished with exit code 0

三、时间处理

rng = pd.date_range('20200601', periods=600, freq='s')
s = pd.Series(np.random.randint(0, 500, len(rng)), index=rng)
print(s.resample('2Min', how='sum'))  #每两分钟进行采样

2020-06-01 00:00:00    32180
2020-06-01 00:02:00    30638
2020-06-01 00:04:00    31144
2020-06-01 00:06:00    28279
2020-06-01 00:08:00    29429
Freq: 2T, dtype: int32

to_timestamp()：将时间序列转化成时间日期的格式

pd.Timestamp('20181020') - pd.Timestamp('20180920')：时间运算

pd.Timestamp('20181020')  + pd.Timedelta(days = 5)：加上五天后的时间

四、类别数据

df = pd.DataFrame({'id': [1, 2, 3, 4, 5, 6], "raw_grade": ['a', 'b', 'b', 'a', 'a', 'd']})
df['grade'] = df.raw_grade.astype('category')  #再给df加一个 raw_grade相同的列

  id raw_grade grade
0   1         a     a
1   2         b     b
2   3         b     b
3   4         a     a
4   5         a     a
5   6         d     d

print(df.grade.cat.categories) #查看类别

df.grade.cat.categories = ['very good', 'good', 'bad']  #对类别重命名
print(df)

   id raw_grade      grade
0   1         a  very good
1   2         b       good
2   3         b       good
3   4         a  very good
4   5         a  very good
5   6         d        bad

df.sort_values(by='grade', ascending=True) #进行排序

s = pd.Series(np.random.randn(1000), index = pd.date_range('20000101', periods=1000))
s = s.cumsum()  #对数据进行求和
plt.plot(s)   #画出图像
plt.show()

五、读写数据

df = pd.DataFrame(np.random.randn(100,4), columns=list('ABCD'))
df.to_csv('data.csv')  #写入到磁盘里面
pd.read_csv('data.csv')  #读数据