Seaborn.countplot ：按数量排序，也按类别排序？

Posted 2023-03-12

【中文标题】Seaborn.countplot ：按数量排序，也按类别排序？

【英文标题】：Seaborn.countplot : order categories by count, also by category?

【发布时间】：2019-08-20 11:25:24

【问题描述】：

所以我了解如何对条形图进行排序（即here）。我找不到的是如何按子类别之一对条形图进行排序。

例如，给定以下数据框，我可以获得条形图。但我想做的是，按Type 的Classic 将其从大到小排序。

import pandas as pd

test_df = pd.DataFrame([
['Jake',    38, 'MW',   'Classic'],
['John',    38,'NW',    'Classic'],
['Sam', 34, 'SE',   'Classic'],
['Sam', 22, 'E' ,'Classic'],
['Joe', 43, 'ESE2', 'Classic'],
['Joe', 34, 'MTN2', 'Classic'],
['Joe', 38, 'MTN2', 'Classic'],
['Scott',   38, 'ESE2', 'Classic'],
['Chris',   34, 'SSE1', 'Classic'],
['Joe', 43, 'S1',   'New'],
['Paul',    34, 'NE2',  'New'],
['Joe', 38, 'MC1',  'New'],
['Joe', 34, 'NE2',  'New'],
['Nick',    38, 'MC1',  'New'],
['Al',  38, 'SSE1', 'New'],
['Al',  34, 'ME',   'New'],
['Al',  34, 'MC1',  'New'],
['Joe', 43, 'S1',   'New']], columns = ['Name','Code_A','Code_B','Type'])


import seaborn as sns
sns.set(style="darkgrid")
palette ="Classic":"#FF9999","New":"#99CC99"


g = sns.countplot(y="Name",
                  palette=palette,
                  hue="Type",
                  data=test_df)

所以而不是：

'Joe' 会在顶部，然后是'Sam'，等等。

【参考方案1】：

添加order 参数。使用pandas.crosstab 和sort_values 获得：

import pandas as pd

test_df = pd.DataFrame([
['Jake',    38, 'MW',   'Classic'],
['John',    38,'NW',    'Classic'],
['Sam', 34, 'SE',   'Classic'],
['Sam', 22, 'E' ,'Classic'],
['Joe', 43, 'ESE2', 'Classic'],
['Joe', 34, 'MTN2', 'Classic'],
['Joe', 38, 'MTN2', 'Classic'],
['Scott',   38, 'ESE2', 'Classic'],
['Chris',   34, 'SSE1', 'Classic'],
['Joe', 43, 'S1',   'New'],
['Paul',    34, 'NE2',  'New'],
['Joe', 38, 'MC1',  'New'],
['Joe', 34, 'NE2',  'New'],
['Nick',    38, 'MC1',  'New'],
['Al',  38, 'SSE1', 'New'],
['Doug',    34, 'ME',   'New'],
['Fred',    34, 'MC1',  'New'],
['Joe', 43, 'S1',   'New']], columns = ['Name','Code_A','Code_B','Type'])


import seaborn as sns
sns.set(style="darkgrid")
palette ="Classic":"#FF9999","New":"#99CC99"

order = pd.crosstab(test_df.Name, test_df.Type).sort_values('Classic', ascending=False).index
g = sns.countplot(y="Name",
                  palette=palette,
                  hue="Type",
                  data=test_df,
                  order=order
                 )

【讨论】：

天哪。我在想它要复杂得多。好的。明白了。

我接受这个答案，因为它正是我想要的。但只是好奇，这将按“经典”排序。如果我想让它先按“新”排序怎么办（我调整了上面的测试数据）

实际上更新了我的答案，误读了关于它是按“经典”计数排序的部分

哦，完美！再次感谢

【参考方案2】：

import pandas as pd

test_df = pd.DataFrame([
['Jake',    38, 'MW',   'Classic'],
['John',    38,'NW',    'Classic'],
['Sam', 34, 'SE',   'Classic'],
['Sam', 22, 'E' ,'Classic'],
['Joe', 43, 'ESE2', 'Classic'],
['Joe', 34, 'MTN2', 'Classic'],
['Joe', 38, 'MTN2', 'Classic'],
['Scott',   38, 'ESE2', 'Classic'],
['Chris',   34, 'SSE1', 'Classic'],
['Joe', 43, 'S1',   'New'],
['Paul',    34, 'NE2',  'New'],
['Joe', 38, 'MC1',  'New'],
['Joe', 34, 'NE2',  'New'],
['Nick',    38, 'MC1',  'New'],
['Al',  38, 'SSE1', 'New'],
['Al',  34, 'ME',   'New'],
['Al',  34, 'MC1',  'New'],
['Joe', 43, 'S1',   'New']], columns = ['Name','Code_A','Code_B','Type'])


import seaborn as sns
sns.set(style="darkgrid")
palette ="Classic":"#FF9999","New":"#99CC99"

sb.countplot(y = 'Name', hue='Type', data=test_df, 
order=test_df['Name'].value_counts().index)