python - 如何按索引值对 Pandas 数据框的列进行排序

Question

是否有可能根据特定索引的值对 pandas df 的列进行排序？

price / time           load_1 load_2 load_3 load_4
price                    50,    68,    23,    12
2018-01-01 00:00:00      12,    65,    37,     8
2018-01-01 00:15:00      13,    54,   112,     6
2018-01-01 00:30:00      58,     12,   96,     4

(第一个指标是能源消耗价格，后面几行代表能源数量，列数不定)

所以它基本上看起来像:

price / time           load_2 load_1 load_3 load_4
price                    68,    50,    23,    12
2018-01-01 00:00:00      65,    12,    37,     8
2018-01-01 00:15:00      54,    13,   112,     6
2018-01-01 00:30:00      12,    58,    96,     4

按价格指数按升序或降序对列进行排序。

Answer 1

如果索引名称是 price 则使用 sort_values axis=1:

print (df.columns)
Index(['load_1', 'load_2', 'load_3', 'load_4'], dtype='object')

df = df.sort_values('price', axis=1, ascending=False)
print (df)
                     load_2  load_1  load_3  load_4
price / time                                       
price                    68      50      23      12
2018-01-01 00:00:00      65      12      37       8
2018-01-01 00:15:00      54      13     112       6
2018-01-01 00:30:00      12      58      96       4

如果 MultiIndex 在列中使用 DataFrame.sort_index :

print (df.columns)
MultiIndex(levels=[['load_1', 'load_2', 'load_3', 'load_4'], ['12', '23', '50', '68']],
           labels=[[0, 1, 2, 3], [2, 3, 1, 0]],
           names=['price / time', 'price'])

df = df.sort_index(axis=1, level=1, ascending=False)
print (df)
price / time        load_2 load_1 load_3 load_4
price                   68     50     23     12
2018-01-01 00:00:00     65     12     37      8
2018-01-01 00:15:00     54     13    112      6
2018-01-01 00:30:00     12     58     96      4

还有一个问题是必须将第二级转换为整数:

a = df.columns.get_level_values(0)
b = df.columns.get_level_values(1).astype(int)

df.columns = pd.MultiIndex.from_arrays([a,b], names=df.columns.names)

print (df.columns)
MultiIndex(levels=[['load_1', 'load_2', 'load_3', 'load_4'], [12, 23, 50, 68]],
           labels=[[0, 1, 2, 3], [2, 3, 1, 0]],
           names=['price / time', 'price'])

df = df.sort_index(axis=1, level=1, ascending=False)
print (df)
price / time        load_2 load_1 load_3 load_4
price                   68     50     23     12
2018-01-01 00:00:00     65     12     37      8
2018-01-01 00:15:00     54     13    112      6
2018-01-01 00:30:00     12     58     96      4