Python常用库之pandas

1. 作用

可以处理关系型或带标签的数据，方便对数据进行处理、抽取和操作；

2. 数据结构

一维：Series

多维：DataFrame

索引：Index

3. Series使用

  3.1 数据定义

# 1.1 普通定义
s1 = pd.Series([1,2,3,4],index=['a','b','c','d'])
print(s1)
# 1.2 numpy数组定义
arr1 = np.arange(0,5,1)
s2 = pd.Series(arr1)
print(s2)
# 1.3 字典定义
dict1 = {'red':200,'blue':1000,'yellow':500}
s3 = pd.Series(dict1)
print(s3)
'''
a    1
b    2
c    3
d    4
dtype: int64
0    0
1    1
2    2
3    3
4    4
dtype: int32
blue      1000
red        200
yellow     500
dtype: int64
'''

  3.2 索引方式(筛选)

print(s1[0:2])
print(s3['blue'])
print(s3[s3>8])
'''
a    1
b    2
dtype: int64
1000
blue      1000
red        200
yellow     500
dtype: int64

'''

  3.3 赋值(修改)

s3['blue'] = 1
print(s3)
'''
blue        1
red       200
yellow    500
dtype: int64
'''

  3.4 常用函数

# 4. 常用函数
s4 = pd.Series([1,2,3,1,4,2])
print(s4.unique())
print(s4.value_counts())
print(s4.isin([0,3]))
print(s4.isnull())
'''
[1 2 3 4]
2    2
1    2
4    1
3    1
dtype: int64
0    False
1    False
2     True
3    False
4    False
5    False
dtype: bool
0    False
1    False
2    False
3    False
4    False
5    False
dtype: bool

'''

4. DataFrame使用

4.1 DataFrame定义

# 1.1 字典形式
data1 = {'color':['blue','red'],'object':['ball','pen'],'price':[1,2]}
# 定义列名
data1 = pd.DataFrame(data1)
print(data1)
# 1.2 嵌套字典
data2 = {'red':{2012:22,2013:33},
         'white':{2011:13,2012:22}}
data2 = pd.DataFrame(data2)
print(data2)
'''
color object  price
0  blue   ball      1
1   red    pen      2
       red  white
2011   NaN   13.0
2012  22.0   22.0
2013  33.0    NaN
'''

4.2 数据索引

print(data1['color'])#按列取数据
print(data1.ix[1:,1:])#取对应范围的值
print("--------")
print(data1.loc[:,'color'])#loc：用列标签(index范围和column范围)选取数据(列名)
print('---------')
print(data1.iloc[:,1])#iloc：用位置(index范围和column范围)选取数据
'''
0    blue
1     red
Name: color, dtype: object
  object  price
1    pen      2
--------
0    blue
1     red
Name: color, dtype: object
---------
0    ball
1     pen
Name: object, dtype: object

'''

5. Index使用

 

# coding=utf-8
import pandas as pd
ser = pd.Series([1,5,3,4],index=['a','e','c','d'])
print(ser)
print(ser.index)  # 索引项
print(ser.idxmin(),ser.idxmax())  # 最小、最大索引
ser.reindex(['1','2','3','4'])
print(ser.index)  # 更换索引
print(ser.sort_values)  #
print(ser.drop(['a']))
'''
a    1
e    5
c    3
d    4
dtype: int64
Index(['a', 'e', 'c', 'd'], dtype='object')
a e
Index(['a', 'e', 'c', 'd'], dtype='object')
<bound method Series.sort_values of a    1
e    5
c    3
d    4
dtype: int64>
e    5
c    3
d    4
dtype: int64
'''

6. 函数映射

import pandas as pd
# pandas映射函数
f = lambda x:x.max()-x.min()
data3 = pd.DataFrame([[1,2,3,4],[5,6,7,8]])
data4 = data3.apply(f,axis=1)  # 1为行，0为列

print(data4)
data5 = data3.apply(f)
print(data5)
'''
0    3
1    3
dtype: int64
0    4
1    4
2    4
3    4
dtype: int64
'''

7. 排序

import pandas as pd
import numpy as np

# 0.定义数据
arr1 = np.arange(0,5,1)
data1 = pd.DataFrame(arr1,index=['a','c','b','d','e'],columns=['A'])
print(data1)
# 1.根据索引排序(增序或降序)
data1_index = data1.sort_index(ascending=False)
print(data1_index)
# 2.根据值排序
data1_values = data1.sort_values(by='A')
print(data1_values)
# 3.排位次
data1_ranking = data1.rank()
print(data1_ranking)
'''
   A
a  0
c  1
b  2
d  3
e  4
   A
e  4
d  3
c  1
b  2
a  0
   A
a  0
c  1
b  2
d  3
e  4
     A
a  1.0
c  2.0
b  3.0
d  4.0
e  5.0

'''

8. 相关性和协方差

#相关性和协方差
import pandas as pd
import numpy as np
seq1 = pd.Series([1,2,3,4],[2003,2004,2005,2006])
seq2 = pd.Series([1,2,3,4],[2003,2004,2005,2007])
seq3 = pd.Series([1,2],[2003,2004])
arr1 = np.arange(1,5,1).reshape((2,2))
data1 = pd.DataFrame(arr1)
print(seq1.corr(seq2))  # 相关度
print(seq1.cov(seq2))  # 协方差
print(data1.corr())
print(data1.cov())
print(data1.corrwith(seq3))
'''
1.0
1.0
     0    1
0  1.0  1.0
1  1.0  1.0
     0    1
0  2.0  2.0
1  2.0  2.0
0   NaN
1   NaN
dtype: float64
'''

9. Nan数据

# coding=utf-8
import pandas as pd
import numpy as np
# 对Nan数据处理
data1 = pd.DataFrame([[6,np.nan,10]],columns=['ball','mug','pen'])
print(pd.DataFrame([[6,np.nan,10]]).shape)  # (1, 3)
print(pd.DataFrame([6,np.nan,10]).shape)  # (3, 1)
# 1.1 为nan赋值
#data1['mug'] = 0
#print(data1)
# 1.2 过滤Nan(过滤使用notnull()针对于series类型)
ser1 = pd.Series([6,np.nan,10],index=['ball','mug','pen'])
print(ser1.notnull())
data2_tep = ser1[ser1.notnull()]
print(data2_tep)
print('-------')
print(data1.dropna(axis=1, how='any'))  # all删除行列均Nan的值，any删除所有的Nan值
# 1.3 填充Nan
data2 = data1.fillna(0)  # 填充Nan为0
print(data2)
'''
(1, 3)
(3, 1)
ball     True
mug     False
pen      True
dtype: bool
ball     6.0
pen     10.0
dtype: float64
-------
   ball  pen
0     6   10
   ball  mug  pen
0     6  0.0   10
'''

10. 等级索引

在单轴上创建多级索引，实现二维情况下操作多维数据

# 4.等级索引
print('-------------')
ser1 = pd.Series(np.random.rand(8),index=[['white','white','white','blue','blue','red','red','red'],['up','down','right','up','down','up','down','left']])
print(ser1['white'])
'''
-------------
up       0.821285
down     0.969028
right    0.376834
dtype: float64
'''