基于 groupby 和 binning 将数据帧拆分为多个数据帧

Question

我在 pandas 中有一个数据框，其中包含我想根据它们的 id（'square'）分组的信息。我想获得每个组的平均亮度，并基于这个平均亮度我想将数据帧分为 4 个类别，并获得 4 个输出数据帧。

示例数据框：

squares = pd.DataFrame({'square': {0: 1.0, 1: 1.0, 2: 2.0, 3: 2.0, 4: 5.0, 5: 6.0, 6: 7.0, 7: 8.0},
                    'time': {0: 1.0, 1: 2.0, 2: 1.0, 3: 2.0, 4: 3.0, 5: 3.0, 6: 4.0, 7: 5.0 },
                    'x': {0: 243, 1: 293, 2: 189, 3: 189, 4: 176, 5: 374, 6: 111, 7: 239},
                    'y': {0: 233, 1: 436, 2: 230, 3: 233, 4: 203, 5: 394, 6: 171, 7: 284}, 
                    'brightness': {0: 1000, 1: 1200, 2: 4000, 3: 5000, 4: 2000, 5: 8000, 6: 1300, 7: 4300 }})

squares = squares.set_index('time')
squares


      brightness     square     x     y 
time
1.0     1000          1.0       243   233
2.0     1200          1.0       293   436
1.0     4000          2.0       189   230
2.0     5000          2.0       189   233
3.0     2000          5.0       176   203
3.0     6000          6.0       374   394 
4.0     1300          7.0       111   171
5.0     4300          8.0       239   284

期望的最终结果：

squares_1

      brightness     square     x     y 
time
1.0     1000          1.0       243   233
2.0     1200          1.0       293   436
3.0     2000          5.0       176   203
4.0     1300          7.0       111   171


squares_2

NaN


squares_3

      brightness     square     x     y 
time
1.0     4000          2.0       189   230
2.0     5000          2.0       189   233
5.0     4300          8.0       239   284


squares_4

      brightness     square     x     y 
time
3.0     6000          6.0       374   394 

我从以下开始：

import pandas as pd
import numpy as np
from matplotlib import pyplot as plt

avg = squares.groupby('square')['brightness'].mean()
n, bins, patches = plt.hist(avg, bins = 4)
inds = np.digitize(avg, bins)

我不太确定如何继续。任何帮助表示赞赏！

Answer 1

您可以使用GroupBy.transform 和mean 来生成与原始DataFrame 相同大小的新Series，然后由cut 分箱并最后创建dictionary of DataFrames：

squares = squares.set_index('time')

labs = [f'squares_{x+1}' for x in range(4)]
g = pd.cut(squares.groupby('square')['brightness'].transform('mean'), bins=4, labels=labs)
print (g)
time
1.0    squares_1
2.0    squares_1
1.0    squares_2
2.0    squares_2
3.0    squares_1
3.0    squares_4
4.0    squares_1
5.0    squares_2
Name: brightness, dtype: category
Categories (4, object): [squares_1 < squares_2 < squares_3 < squares_4]

dfs = dict(tuple(squares.groupby(g)))

---

print (dfs)
{'squares_1':       square    x    y  brightness
time                              
1.0      1.0  243  233        1000
2.0      1.0  293  436        1200
3.0      5.0  176  203        2000
4.0      7.0  111  171        1300, 'squares_2':       square    x    y  brightness
time                              
1.0      2.0  189  230        4000
2.0      2.0  189  233        5000
5.0      8.0  239  284        4300, 'squares_3': Empty DataFrame
Columns: [square, x, y, brightness]
Index: [], 'squares_4':       square    x    y  brightness
time                              
3.0      6.0  374  394        8000}

---

print (dfs['squares_1'])
      square    x    y  brightness
time                              
1.0      1.0  243  233        1000
2.0      1.0  293  436        1200
3.0      5.0  176  203        2000
4.0      7.0  111  171        1300

print (dfs['squares_2'])
      square    x    y  brightness
time                              
1.0      2.0  189  230        4000
2.0      2.0  189  233        5000
5.0      8.0  239  284        4300

print (dfs['squares_3'])
Empty DataFrame
Columns: [square, x, y, brightness]
Index: []

print (dfs['squares_4'])
      square    x    y  brightness
time                              
3.0      6.0  374  394        8000