DataFrame 是 Pandas 的另一个重要对象数据类型，其属于二维数组的形式，类似于Excel的一张表。可以同时存储数据分析中的多个变量，是进行统计分析和机器学习的基本数据结构。
1 创建DataFrame
创建空DataFrame：
>>> import pandas as pd
>>> import numpy as np # pandas基于numpy开发，所以需要导入numpy库
>>> df = pd.DataFrame()
>>> print(df)
Empty DataFrame
Columns: []
Index: []
列表创建DataFrame：
>>> list1=[24,37,18,27,33,40,16]
>>> df1=pd.DataFrame(list1) # 按列创建DataFrame
>>> df1
0
0 24
1 37
2 18
3 27
4 33
5 40
6 16
>>> list1_1=[[24,37,18,27,33,40,16]]
>>> df1_1=pd.DataFrame(list1_1) # 按行创建DataFrame
>>> df1_1
0 1 2 3 4 5 6
0 24 37 18 27 33 40 16
>>> pd.DataFrame(list1,index=['a','b','c','d','e','f','g'],columns=['age'])
age
a 24
b 37
c 18
d 27
e 33
f 40
g 16
>>> list2=[[24,'M'],[37,'F'],[18,'F'],[27,'M'],[33,'F'],[40,'M']]
>>> pd.DataFrame(list2,columns=['age','sex'])
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
>>> pd.DataFrame(list(map(list,zip(*list2))),index=['age','sex']) # 先将list2转置后再创建DataFrame，从而实现按照列生成DataFrame。
0 1 2 3 4 5
age 24 37 18 27 33 40
sex M F F M F M
向量创建DataFrame：
>>> array1=np.array([24,37,18,27,33,40,16])
>>> pd.DataFrame(array1,index=['a','b','c','d','e','f','g'],columns=['age'])
age
a 24
b 37
c 18
d 27
e 33
f 40
g 16
>>> array2=np.array([[24,'M'],[37,'F'],[18,'F'],[27,'M'],[33,'F'],[40,'M']])
>>> pd.DataFrame(array2,columns=['age','sex'])
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
>>> pd.DataFrame(array2.T,index=['age','sex']) # 可以通过array转置的方式按照列创建DataFrame
0 1 2 3 4 5
age 24 37 18 27 33 40
sex M F F M F M
字典创建DataFrame：
>>> dict1 = {'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',]}
>>> pd.DataFrame(dict1)
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
Series创建DataFrame：
>>> s1=pd.Series([24,37,18,27,33,40])
>>> s2=pd.Series(['M','F','F','M','F','M'])
>>> pd.DataFrame([s1,s2],index=['age','sex']) # 默认是按照行来创建DataFrame！
0 1 2 3 4 5
age 24 37 18 27 33 40
sex M F F M F M
>>> pd.DataFrame({'age':s1,'sex':s2}) # 可以将Series添加到字典，从而按照列来创建DataFrame！
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
index的修改与重置：
>>> df1_1=df1_2=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',]})
>>> df1_1
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
>>> df1_1.index=df1_1['age']
>>> df1_1
age sex
age
24 24 M
37 37 F
18 18 F
27 27 M
33 33 F
40 40 M
>>> df1_2.set_index('age', drop= True, inplace=True)
>>> df1_2
sex
age
24 M
37 F
18 F
27 M
33 F
40 M
>>> df1_1.reset_index(drop=True)
sex
0 M
1 F
2 F
3 M
4 F
5 M
df1_1.reset_index(drop=False)
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
2 DataFrame数据的索引、修改和删除
索引和修改：
>>> df2=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',]})
>>> df2['age']
0 24
1 37
2 18
3 27
4 33
5 40
Name: age, dtype: int64
>>> df2[['age','sex']]
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
>>> df2.loc[1,:] # 按照行的index进行索引
age 37
sex F
>>> df2.iloc[1,:] # 按照行的位置次序进行索引，注意行的index可以变，但是行的位置次序不会变。
age 37
sex F
Name: 1, dtype: object
>>> df2.loc[1:3,'age':'sex'] # 根据行和列的index同时索引
age sex
1 37 F
2 18 F
3 27 M
>>> df2.iloc[1:4,0:2] # 根据行和列的位置次序同时索引
age sex
1 37 F
2 18 F
3 27 M
>>> df2.loc[4,'sex']
'F'
>>> df2.iloc[4,1]
'F'
>>> df2.loc[1,'age']=38
>>> df2
age sex
0 24 M
1 38 F
2 18 F
3 27 M
4 33 F
5 40 M
>>> df2.iloc[1,0]=37
>>> df2
age sex
0 24 M
1 37 F
2 18 F
3 27 M
4 33 F
5 40 M
添加：
>>> df2=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',]})
>>> df2['age']
0 24
1 37
2 18
3 27
4 33
5 40
Name: age, dtype: int64
>>> df2.insert(1,column='name',value=['denise','maggie','eileen','cyril','julian','colin'])
>>> df2
age name sex
0 24 denise M
1 37 maggie F
2 18 eileen F
3 27 cyril M
4 33 julian F
5 40 colin M
>>> df2.loc[len(df2.index)] = [20, 'jack', 'M'] # 最后一行插入行
>>> df2
age name sex
0 24 denise M
1 37 maggie F
2 18 eileen F
3 27 cyril M
4 33 julian F
5 40 colin M
6 20 jack M
删除：
接着上面的df2操作
>>> df2.drop('sex',axis=1) # 删除1列，如果需要修改df2的值需要加参数`inplace='True'`。
age name
0 24 denise
1 37 maggie
2 18 eileen
3 27 cyril
4 33 julian
5 40 colin
6 20 jack
>>> df2.drop(['age','sex'],axis=1) # 删除多列，如果需要修改df2的值需要加参数`inplace='True'`。
name
0 denise
1 maggie
2 eileen
3 cyril
4 julian
5 colin
6 jack
>>> df2.drop(6,axis=0) # 删除1行，如果需要修改df2的值需要加参数`inplace='True'`。
age name sex
0 24 denise M
1 37 maggie F
2 18 eileen F
3 27 cyril M
4 33 julian F
5 40 colin M
>>> df2.drop(range(2,4),axis=0) # 删除多行，如果需要修改df2的值需要加参数`inplace='True'`。
age name sex
0 24 denise M
1 37 maggie F
4 33 julian F
5 40 colin M
6 20 jack M
3 获取DataFrame的属性
>>> df3=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',],'name':['denise','maggie','eileen','cyril','julian','colin']})
>>> df3.T
0 1 2 3 4 5
age 24 37 18 27 33 40
sex M F F M F M
name denise maggie eileen cyril julian colin
>>> df3.axes
[RangeIndex(start=0, stop=6, step=1), Index(['age', 'sex', 'name'], dtype='object')]
>>> df3.ndim
2
>>> df3.shape
(6, 3)
>>> df3.values
...
array([[24, 'M', 'denise'],
[37, 'F', 'maggie'],
[18, 'F', 'eileen'],
[27, 'M', 'cyril'],
[33, 'F', 'julian'],
[40, 'M', 'colin']], dtype=object)
>>> df3.head(3)
age sex name
0 24 M denise
1 37 F maggie
2 18 F eileen
>>> df3.tail(2)
age sex name
4 33 F julian
5 40 M colin
4 DataFrame中的函数操作
>>> df3.sort_values(by='age',axis=0)
age sex name
2 18 F eileen
0 24 M denise
3 27 M cyril
4 33 F julian
1 37 F maggie
5 40 M colin
>>> df3.sort_index()
age sex name
0 24 M denise
1 37 F maggie
2 18 F eileen
3 27 M cyril
4 33 F julian
5 40 M colin
>>> df3.replace('M','F')
age sex name
0 24 F denise
1 37 F maggie
2 18 F eileen
3 27 F cyril
4 33 F julian
5 40 F colin
>>> df3.drop_duplicates('sex')
age sex name
0 24 M denise
1 37 F maggie
>>> df4=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',],'weight':[76,56,49,84,61,66]})
>>> df4
age sex weight
0 24 M 76
1 37 F 56
2 18 F 49
3 27 M 84
4 33 F 61
5 40 M 66
>>> df4.mean(axis=0)
age 29.833333
weight 65.333333
dtype: float64
>>> df4.mean(axis=1)
0 50.0
1 46.5
2 33.5
3 55.5
4 47.0
5 53.0
dtype: float64
>>> df4.describe()
age weight
count 6.000000 6.000000
mean 29.833333 65.333333
std 8.328665 12.925427
min 18.000000 49.000000
25% 24.750000 57.250000
50% 30.000000 63.500000
75% 36.000000 73.500000
max 40.000000 84.000000
>>> df4.describe(include='object')
sex
count 6
unique 2
top M
freq 3
5 合并DataFrame对象
按照索引合并：
>>> df5_1=pd.DataFrame({'age': [24,37,18,27,33,40], 'sex': ['M','F','F','M','F','M',]})
>>> df5_2=pd.DataFrame({'name': ['james','denise','eileen','kobe','julian','colin'],'weight':[76,56,49,84,61,66]},index=[2,3,4,5,6,7])
>>> df5_3=pd.DataFrame({'name': ['denise','maggie','eileen','cyril','julian','colin'],'height':[164,178,152,160,171,166]})
>>> pd.concat([df5_1,df5_2],axis=1)
age sex name weight
0 24.0 M NaN NaN
1 37.0 F NaN NaN
2 18.0 F james 76.0
3 27.0 M denise 56.0
4 33.0 F eileen 49.0
5 40.0 M kobe 84.0
6 NaN NaN julian 61.0
7 NaN NaN colin 66.0
>>> pd.concat([df5_1,df5_2],axis=1,join='inner')
age sex name weight
2 18 F james 76
3 27 M denise 56
4 33 F eileen 49
5 40 M kobe 84
>>> pd.concat([df5_2,df5_3])
name weight height
2 james 76.0 NaN
3 denise 56.0 NaN
4 eileen 49.0 NaN
5 kobe 84.0 NaN
6 julian 61.0 NaN
7 colin 66.0 NaN
0 denise NaN 164.0
1 maggie NaN 178.0
2 eileen NaN 152.0
3 cyril NaN 160.0
4 julian NaN 171.0
5 colin NaN 166.0
>>> pd.concat([df5_2,df5_3],join='inner')
name
2 james
3 denise
4 eileen
5 kobe
6 julian
7 colin
0 denise
1 maggie
2 eileen
3 cyril
4 julian
5 colin
>>> pd.concat([df5_2,df5_3],join='inner',ignore_index=True)
name
0 james
1 denise
2 eileen
3 kobe
4 julian
5 colin
6 denise
7 maggie
8 eileen
9 cyril
10 julian
11 colin
按照DataFrame的列进行合并：
在实际数据分析时，往往需要从多个表中导入数据并根据关键词进行合并，这时可以使用merge来进行合并：
>>> df5_2=pd.DataFrame({'name': ['james','denise','eileen','kobe','julian','colin'],'weight':[76,56,49,84,61,66]},index=[2,3,4,5,6,7])
>>> df5_3=pd.DataFrame({'name': ['denise','maggie','eileen','cyril','julian','colin'],'height':[164,178,152,160,171,166]})
>>> df5_2.merge(df5_3,on='name')
name weight height
0 denise 56 164
1 eileen 49 152
2 julian 61 171
3 colin 66 166