在 Pandas 中使用 Merge、Join 、Concat合并数据的效率对比

deephub

AI方向文章，看头像就知道，这里都是"干"货

在 Pandas 中有很多种方法可以进行DF的合并。本文将研究这些不同的方法，以及如何将它们应用到我们的数据中。

合并DF

Pandas 使用 .merge() 方法来执行合并。

import pandas as pd  
# a dictionary to convert to a dataframe 
data1 = {'identification': ['a', 'b', 'c', 'd'], 
        'Customer_Name':['King', 'West', 'Adams', 'Mercy'],          'Category':['furniture', 'Office Supplies', 'Technology', 'R_materials'],}  
# our second dictionary to convert to a dataframe  
data2 = {'identification': ['a', 'b', 'c', 'd'], 
        'Class':['First_Class', 'Second_Class', 'Same_day', 'Standard Class'],  
       'Age':[60, 30, 40, 50]}  
# Convert the dictionary into DataFrame   
df1 = pd.DataFrame(data1) 
df2 = pd.DataFrame(data2)

运行我们的代码后，有两个 DataFrame，如下所示。

identification Customer_Name         Category 
0              a          King        furniture 
1              b          West  Office Supplies 
2              c         Adams       Technology 
3              d         Mercy      R_materials  
 identification           Class  Age 
0              a     First_Class   60 
1              b    Second_Class   30 
2              c        Same_day   40 
3              d  Standard Class   50

使用 merge() 函数进一步合并。

# using .merge() function  
new_data = pd.merge(df1, df2, on='identification')

这产生了下面的新数据；

  identification Customer_Name  Category     Class            Age 
0     a           King          furniture     First_Class     60 
1     b           West         Office Supplies Second_Class   30 
2     c           Adams         Technology      Same_day      40 
3     d           Mercy         R_materials  Standard Class   50

.join() 方法也可以将不同索引的 DataFrame 组合成一个新的 DataFrame。我们可以使用参数‘on’参数指定根据哪列进行合并。

让我们看看下面的例子，我们如何将单索引 DataFrame 与多索引 DataFrame 连接起来；

import pandas as pd  
# a dictionary to convert to a dataframe 
data1 = { 
       'Customer_Name':['King', 'West', 'Adams'],  
      'Category':['furniture', 'Office Supplies', 'Technology'],}  7     
# our second dictionary to convert to a dataframe  
data2 = { 
       'Class':['First_Class', 'Second_Class', 'Same_day', 'Standard Class'],  
      'Age':[60, 30, 40, 50]}  
# Convert the dictionary into DataFrame   
Ndata = pd.DataFrame(data1, index=pd.Index(['a', 'b', 'c'], name='identification')) 
index = pd.MultiIndex.from_tuples([('a', 'x0'), ('b', 'x1'), 
                                 ('c', 'x2'), ('c', 'x3')], 
                                 names=['identification', 'x']) 19   
# Convert the dictionary into DataFrame   
Ndata2 = pd.DataFrame(data2, index= index) 
print(Ndata, "\n\n", Ndata2) 
# joining singly indexed with 
# multi indexed 
result = Ndata.join(Ndata2, how='inner')

我们的结果如下所示；

               Customer_Name       Category      Class       Age 
identification x                                                      3 a         x0       King        furniture     First_Class      60 
b         x1       West     Office Supplies    Second_Class   30 
c         x2       Adams       Technology       Same_day      40 
          x3       Adams       Technology  Standard Class     50

连接DF

Pandas 中concat() 方法在可以在垂直方向（axis=0）和水平方向（axis=1）上连接 DataFrame。我们还可以一次连接两个以上的 DataFrame 或 Series。

让我们看一个如何在 Pandas 中执行连接的示例；

import pandas as pd   
# a dictionary to convert to a dataframe  
data1 = {'identification': ['a', 'b', 'c', 'd'],  
        'Customer_Name':['King', 'West', 'Adams', 'Mercy'],   
       'Category':['furniture', 'Office Supplies', 'Technology', 'R_materials'],}   
# our second dictionary to convert to a dataframe   
data2 = {'identification': ['a', 'b', 'c', 'd'],  
        'Class':['First_Class', 'Second_Class', 'Same_day', 'Standard Class'],   
       'Age':[60, 30, 40, 50]}   
# Convert the dictionary into DataFrame    
df1 = pd.DataFrame(data1)  
df2 = pd.DataFrame(data2)   
#perform concatenation here based on horizontal axis  
new_data = pd.concat([df1, df2], axis=1)  
print(new_data)

这样就获得了新的 DataFrame ：

 identification Customer_Name         Category identification  \ 
0              a          King        furniture              a    3 1              b          West  Office Supplies              b    4 2              c         Adams       Technology              c    5 3              d         Mercy      R_materials              d