pandas数据拼接

一 前言

pandas数据拼接有可能会用到，比如出现重复数据，需要合并两份数据的交集，并集就是个不错的选择，知识追寻者本着技多不压身的态度蛮学习了一下下；

公众号：知识追寻者

知识追寻者(Inheriting the spirit of open source, Spreading technology knowledge;)

二 数据拼接

在进行学习数据转换之前，先学习一些数拼接相关的知识

2.1 join()联结

有关merge操作知识追寻者这边不提及，有空可能后面会专门出一篇相关文章，因为其学习方式根SQL的表联结类似，不是几行能说清楚的知识点；

join操作能将 2 个DataFrame 合并为一块，前提是DataFrame 之间的列没有重复；

# -*- coding: utf-8 -*-

import pandas as pd
import numpy as np

data1 = {
    'user' : ['zszxz','craler','rose'],
    'price' : [100, 200, 300],
    'hobby' : ['reading','running','hiking']
}
index1 = ['user1','user2','user3']
frame1  = pd.DataFrame(data1,index1)

data2 = {
    'person' : ['zszxz','craler','rose'],
    'number' : [100, 2000, 3000],
    'activity' : ['swing','riding','climbing']
}
index2 = ['user1','user2','user3']
frame2  = pd.DataFrame(data2,index2)

join = frame1.join(frame2)
print(join)

输出

         user  price    hobby  person  number  activity
user1   zszxz    100  reading   zszxz     100     swing
user2  craler    200  running  craler    2000    riding
user3    rose    300   hiking    rose    3000  climbing

2.2 concat()拼接

使用 concat() 函数能将2个 Series 拼接为一个，默认按行拼接；

ser1 = pd.Series(['111','222',np.NaN])
ser2 = pd.Series(['333','444',np.NaN])
# 默认按行拼接
print(pd.concat([ser1, ser2]))

如果按列拼接则 axis = 1

ser1 = pd.Series(['111','222',np.NaN])
ser2 = pd.Series(['333','444',np.NaN])
# 按列拼接
print(pd.concat([ser1, ser2],axis=1))

输出

     0    1
0  111  333
1  222  444
2  NaN  NaN

更近一步，指定key 参数 输出的数据格式就和 DataFrame 一样

ser1 = pd.Series(['111','222',np.NaN])
ser2 = pd.Series(['333','444',np.NaN])
# 按列拼接
data = pd.concat([ser1, ser2],axis=1, keys=['zszxz', 'rzxx'])
print(data)

输出

  zszxz rzxx
0   111  333
1   222  444
2   NaN  NaN

注 ： DataFrame 的 concat 操作 和 Series 类似；

2.3 combine_first()组合

索引重复时就可以使用combine_first进行拼接

ser1 = pd.Series(['111','222',np.NaN],index=[1,2,3])
ser2 = pd.Series(['333','444',np.NaN,'555'],index=[1,2,3,4])
data = ser1.combine_first(ser2)
print(data)

输出

1    111
2    222
3    NaN
4    555
dtype: object

将Series 位置互换一下，可以看见基准将以 ser2为准；

ser1 = pd.Series(['111','222',np.NaN],index=[1,2,3])
ser2 = pd.Series(['333','444',np.NaN,'555'],index=[1,2,3,4])
data = ser2.combine_first(ser1)
print(data)

输出

1    333
2    444
3    NaN
4    555
dtype: object

2.4 轴转换

准备的数据

# -*- coding: utf-8 -*-

import pandas as pd
import numpy as np

data = {
    'user' : ['zszxz','craler','rose'],
    'price' : [100, 200, 300],
    'hobby' : ['reading','running','hiking']
}
index = ['user1','user2','user3']
frame  = pd.DataFrame(data,index)
print(frame)

输出

         user  price    hobby
user1   zszxz    100  reading
user2  craler    200  running
user3    rose    300   hiking

• stack() 将 列转为行；

# -*- coding: utf-8 -*-

import pandas as pd
import numpy as np

data = {
    'user' : ['zszxz','craler','rose'],
    'price' : [100, 200, 300],
    'hobby' : ['reading','running','hiking']
}
index = ['user1','user2','user3']
frame  = pd.DataFrame(data,index)
print(frame.stack())

输出

user1  user       zszxz
       price        100
       hobby    reading
user2  user      craler
       price        200
       hobby    running
user3  user        rose
       price        300
       hobby     hiking
dtype: object

• 使用 unstack()将 数据结构重新返回

# -*- coding: utf-8 -*-

import pandas as pd
import numpy as np

data = {
    'user' : ['zszxz','craler','rose'],
    'price' : [100, 200, 300],
    'hobby' : ['reading','running','hiking']
}
index = ['user1','user2','user3']
frame  = pd.DataFrame(data,index)
sta = frame.stack()
print(sta.unstack())

输出

         user price    hobby
user1   zszxz   100  reading
user2  craler   200  running
user3    rose   300   hiking