pandas

pandas索引

>import pandas as pd
>s.index.name='alpha'#给一维数组索引定义名字
#行索引
>df.index

#列索引
>df.columns

#索引值有重复项时
>s.index.is_unique #判断是否有重复索引
>s.groupby(s.index).sum()  #将重复项求和

#多级索引
>pd.MultiIndex   
#索引类

分组计算

拆分-应用-合并
拆分：根据什么进行分组
应用：每个分组进行什么样的计算
合并：把每个分组的计算结果合并起来

>df=pd.DataFrame({'key1':['a','a','b','b','a'],'key2':['one','two','one','two','one'],'data1':np.random.randint(1,10,5),'data2':np.random.randint(1,10,5)})
df
Out[10]: 
  key1 key2  data1  data2
0    a  one      4      2
1    a  two      9      1
2    b  one      7      8
3    b  two      9      8
4    a  one      2      2
>df['data1'].groupby(df['key1']).mean()  #也可以多层分组，可在‘df['key1']’后面加上df['key2']
Out[13]: 
key1
a    5
b    8
Name: data1, dtype: int32

>df.groupby('key1').sum()['data1'] 
Out[15]: 
key1
a    15
b    16
Name: data1, dtype: int32

>sum=df.groupby(['key1','key2']).sum()['data1']
sum.unstack()
Out[24]: 
key2  one  two
key1          
a       6    9
b       7    9

#用for循环迭代,分成两个分组
>for name,group in df.groupby('key1'):
    print (name)
    print (group)
    
a
  key1 key2  data1  data2
0    a  one      4      2
1    a  two      9      1
4    a  one      2      2
b
  key1 key2  data1  data2
2    b  one      7      8
3    b  two      9      8

通过字典分组

>df=pd.DataFrame(np.random.randint(1,10,(5,5)),columns=['a','b','c','d','e'],index=['Alice','Bob','Candy','Dark','Emily'])
df.iloc[1,1:3]=np.NaN
df
Out[37]: 
       a    b    c  d  e
Alice  3  2.0  9.0  1  6
Bob    3  NaN  NaN  4  7
Candy  8  3.0  8.0  4  9
Dark   5  8.0  9.0  3  1
Emily  7  7.0  8.0  9  6

#关注分组对于非数字的处理
>mapping={'a':'red','b':'red','c':'blue','d':'orange','e':'blue'}
grouped=df.groupby(mapping,axis=1)
grouped.sum()
Out[46]:              #对NaN当成0处理
       blue  orange   red
Alice  15.0     1.0   5.0
Bob     7.0     4.0   3.0
Candy  17.0     4.0  11.0
Dark   10.0     3.0  13.0
Emily  14.0     9.0  14.0

通过函数分组

>def _group_key(idx):
    print(idx)
    return (idx)
df.groupby(_group_key)
Alice
Bob
Candy
Dark
Emily

聚合运算

#内置聚合函数
>df.groupby('key1').desceibe()

#自定义聚合函数
>grouped=df.groupby('key1')
>def peak_range(s):
>   print type(s)
>   return s.max()-s.min()
>grouped.agg(peak_range)