如何对堆积条形图进行分组？

Question

我正在尝试创建一个分组的堆叠条形图。

目前我有以下数据框：

>>> df
                                       Value                     
Rating                                 1          2         3
Context Parameter                                
Total   1                          43.312347   9.507902  1.580367
        2                          42.862649   9.482205  1.310549
        3                          43.710651   9.430811  1.400488
        4                          43.209559   9.803418  1.349094
        5                          42.541436  10.008994  1.220609
        6                          42.978286   9.430811  1.336246
        7                          42.734164  10.317358  1.606064
User    1                          47.652348  11.138861  2.297702
        2                          47.102897  10.589411  1.848152
        3                          46.853147  10.139860  1.848152
        4                          47.252747  11.138861  1.748252
        5                          45.954046  10.239760  1.448551
        6                          46.353646  10.439560  1.498501
        7                          47.102897  11.338661  1.998002

我希望将每个参数的“总计”和“用户”栏分组在一起。

这是生成的图表df.plot(kind='bar', stacked=True)：

这些条形本身看起来是正确的，但是如何让每个Parameter的Total和User条形图彼此相邻，最好在参数之间留有一些余量？

Answer 1

以下方法允许同时分组和堆叠条。首先，数据框按 排序parameter, context。然后context从索引中取出，为每context, value对创建新列。最后，绘制三个相互重叠的条形图以可视化堆叠的条形图。

import pandas as pd
from matplotlib import pyplot as plt

df = pd.DataFrame(columns=['Context', 'Parameter', 'Val1', 'Val2', 'Val3'],
                  data=[['Total', 1, 43.312347, 9.507902, 1.580367],
                        ['Total', 2, 42.862649, 9.482205, 1.310549],
                        ['Total', 3, 43.710651, 9.430811, 1.400488],
                        ['Total', 4, 43.209559, 9.803418, 1.349094],
                        ['Total', 5, 42.541436, 10.008994, 1.220609],
                        ['Total', 6, 42.978286, 9.430811, 1.336246],
                        ['Total', 7, 42.734164, 10.317358, 1.606064],
                        ['User', 1, 47.652348, 11.138861, 2.297702],
                        ['User', 2, 47.102897, 10.589411, 1.848152],
                        ['User', 3, 46.853147, 10.139860, 1.848152],
                        ['User', 4, 47.252747, 11.138861, 1.748252],
                        ['User', 5, 45.954046, 10.239760, 1.448551],
                        ['User', 6, 46.353646, 10.439560, 1.498501],
                        ['User', 7, 47.102897, 11.338661, 1.998002]])
df.set_index(['Context', 'Parameter'], inplace=True)
df0 = df.reorder_levels(['Parameter', 'Context']).sort_index()

colors = plt.cm.Paired.colors

df0 = df0.unstack(level=-1) # unstack the 'Context' column
fig, ax = plt.subplots()
(df0['Val1']+df0['Val2']+df0['Val3']).plot(kind='bar', color=[colors[1], colors[0]], rot=0, ax=ax)
(df0['Val2']+df0['Val3']).plot(kind='bar', color=[colors[3], colors[2]], rot=0, ax=ax)
df0['Val3'].plot(kind='bar', color=[colors[5], colors[4]], rot=0, ax=ax)

legend_labels = [f'{val} ({context})' for val, context in df0.columns]
ax.legend(legend_labels)

plt.tight_layout()
plt.show()