代码之家 › 专栏 › 技术社区 › natemcintosh

Pandas groupby:将两列视为一列

pandas-groupby dataframe pandas python-3.x python

natemcintosh · 技术社区 · 7 年前

我可以通过将这两列转换为一列元组来执行groupby操作,并按该列进行分组。然而,我的实际数据帧非常大,添加另一列可能会减慢速度。我想知道在熊猫身上是否有更惯用的方法来做到这一点。

In [1]: import pandas as pd                                                                                                                                                                                 
In [2]: import numpy as np                                                                                                                                                                                  
In [3]: key  = np.random.randint(low = 1, high = 20, size = 100) 
   ...: data = np.random.random(size = (100)) 
   ...: d1   = {'key':key, 'time':range(1,101), 'data':data} 
   ...: df1  = pd.DataFrame(d1) 
   ...: print(df1.shape) 
   ...: df1.head()                                                                                                                                                                                                 
(100, 3)
Out[3]: 
    key  time      data
0     3     1  0.778231
1    13     2  0.822494
2     4     3  0.053416
3     8     4  0.894341
4     7     5  0.884310
In [4]: key = range(1,21) 
   ...: lat = np.random.randint(low = 0, high = 90, size = 20) 
   ...: lon = np.random.randint(low = 0, high = 90, size = 20) 
   ...: d2  = {'key':key, 'lat':lat, 'lon':lon} 
   ...: df2 = pd.DataFrame(d2) 
   ...: print(df2.shape) 
   ...: df2.head()                                                                                                                                                                                                 
(20, 3)
Out[4]: 
    key  lat  lon
0     1   36   81
1     2    6   57
2     3   84    4
3     4   61    0
4     5   54   69
In [5]: result = pd.merge(df1, df2).sort_values('time') 
   ...: result.head()                                                                                                                                                                                            
Out[5]: 
    key  time      data  lat  lon
0     3     1  0.778231   84    4
4    13     2  0.822494   12   19
13    4     3  0.053416   61    0
18    8     4  0.894341   49   34
23    7     5  0.884310    8   13

(确保在框中向下滚动以查看的输出。) In [5]

result.groupby(('lat','lon')) 并让熊猫将两列视为一列。有办法做到这一点吗?还是我应该咬紧牙关,创建一个新的数据元组列?

1 回复 | 直到 7 年前

Szymon Maszke 7 年前

在这一点上,我希望能够做到 result.groupby(('lat','lon'))

示例数据:

key  time     data  lat  lon
3     1   0.231000   84    4
4     1   0.832310   22   11
5     1   1.210000   84    4
6     1   3.778231   22   11
8     1  15.450000   84    4

import pandas as pd

for name, group in df.groupby(["lat", "lon"]):
    print("Group indices: {}".format(name))
    print(group)

Group indices: (22, 11)
   key  time      data  lat  lon
1    4     1  0.832310   22   11
3    6     1  3.778231   22   11
Group indices: (84, 4)
   key  time    data  lat  lon
0    3     1   0.231   84    4
2    5     1   1.210   84    4
4    8     1  15.450   84    4

这不正是你想要的吗?还是我误解了什么?