Pandas，避免数据透视表中的层次结构答案

【问题标题】：Pandas, Avoiding hierarchy in pivot tablePandas，避免数据透视表中的层次结构
【发布时间】：2016-04-24 06:46:44
【问题描述】：

我有一个熊猫数据框df，使用以下函数从中生成数据透视表；

def objective2(excel_file):
    df = pd.read_excel(excel_file)

    # WBC cut-offs
    df['WBC_groups'] = pd.cut(df.WBC, [0, 4, 12, 100], 
                             labels=['WBC < 4', 'WBC Normal', 'WBC > 12'])

    df['count'] = 1

    table = df.pivot_table('count', index=['Sex'],
                           columns=['WBC_groups', 'Outcome_at_24'],
                           aggfunc='sum',
                           margins=True, margins_name='Total')

    return table

这会生成下表：

WBC_groups         WBC < 4      WBC Normal      WBC > 12      Total
Outcome_at_24   Alive Died      Alive Died    Alive Died       
Sex                                                            
Female           10.0  2.0       20.0  6.0     14.0  NaN       86.0
Male              3.0  NaN       28.0  3.0     26.0  4.0      111.0
Total            13.0  2.0       48.0  9.0     40.0  4.0      197.0

如何避免列中的层次结构，使表格看起来像这样：

WBC_groups       WBC < 4    WBC Normal   WBC > 12   Alive   Died  Total      
Sex                                                            
Female           10.0          2.0       20.0       6.0     14.0  86.0
Male              3.0          NaN       28.0       3.0     26.0  111.0
Total            13.0          2.0       48.0       9.0     40.0  197.0

注意：表格中的数据并不准确，只是假数据。

【问题讨论】：

输出是否正确？列(WBC > 12), Died 被丢弃？
@jezrael，输出是正确的，但总列取整个数据集，因此根据数据集是正确的。 NaN 值为零。

标签： python pandas

【解决方案1】：

我认为你不能避免层次结构，因为在pivot_table 中使用带有两列的参数列 - WBC_groups 和 Outcome_at_24。

最简单的解决方案是设置新的列名，然后设置drop 列rem：

df.columns = ['WBC < 4', 'WBC Normal', 'WBC > 12', 'Alive', 'Died', 'rem', 'Total']
df = df.drop('rem', axis=1)
print df
        WBC < 4  WBC Normal  WBC > 12  Alive  Died  Total
Sex                                                      
Female     10.0         2.0      20.0    6.0  14.0   86.0
Male        3.0         NaN      28.0    3.0  26.0  111.0
Total      13.0         2.0      48.0    9.0  40.0  197.0

但如果您需要更通用的解决方案：

print df
WBC_groups    WBC < 4      WBC Normal      WBC > 12       Total
Outcome_at_24   Alive Died      Alive Died    Alive Died       
Sex                                                            
Female           10.0  2.0       20.0  6.0     14.0  NaN   86.0
Male              3.0  NaN       28.0  3.0     26.0  4.0  111.0
Total            13.0  2.0       48.0  9.0     40.0  4.0  197.0

cols1 = df.columns.get_level_values('WBC_groups').to_series().drop_duplicates().tolist()
print cols1
['WBC < 4', 'WBC Normal', 'WBC > 12', 'Total']

cols2 = df.columns.get_level_values('Outcome_at_24').to_series().drop_duplicates().tolist()
print cols2
['Alive', 'Died', ' ']

cols = cols1[:-1] + cols2[:2] + ['rem'] + cols1[-1:]
print cols
['WBC < 4', 'WBC Normal', 'WBC > 12', 'Alive', 'Died', 'rem', 'Total']

df.columns = cols

df = df.drop('rem', axis=1)
print df
        WBC < 4  WBC Normal  WBC > 12  Alive  Died  Total
Sex                                                      
Female     10.0         2.0      20.0    6.0  14.0   86.0
Male        3.0         NaN      28.0    3.0  26.0  111.0
Total      13.0         2.0      48.0    9.0  40.0  197.0

【讨论】：