Question

我试图在数据框中创建一个新系列，该系列沿二维方式映射字典，首先匹配键，然后匹配数组中的值。现有系列是日期时间，键匹配是日期，值匹配是小时（因此新系列'dh'）

在这里映射一维数组有一个类似的问题：Adding a new pandas column with mapped value from a dictionary，但是将整个数组映射到每一天。

当前代码：

import pandas as pd

df = pd.DataFrame({'datetime':pd.date_range('1/1/2018','1/4/2018', freq = '1H', closed = 'left')})

day_hour = {1:range(48,0,-2),
            2:range(96,0,-4),
            3:range(120,0,-5) }

df['dh'] = df['datetime'].dt.day.map(day_hour)

输出摘录：

              datetime                                                 dh
0  2018-01-01 00:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
1  2018-01-01 01:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
2  2018-01-01 02:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
3  2018-01-01 03:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
4  2018-01-01 04:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
5  2018-01-01 05:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
6  2018-01-01 06:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
7  2018-01-01 07:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
8  2018-01-01 08:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
9  2018-01-01 09:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
10 2018-01-01 10:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...
11 2018-01-01 11:00:00  [48, 46, 44, 42, 40, 38, 36, 34, 32, 30, 28, 2...

期望的输出：

              datetime   dh
0  2018-01-01 00:00:00   48
1  2018-01-01 01:00:00   46
2  2018-01-01 02:00:00   44
3  2018-01-01 03:00:00   42
4  2018-01-01 04:00:00   40
5  2018-01-01 05:00:00   38
...

Answer 1

如果您的dict创建得很好，则不需要地图

df['dh']=sum(map(list,day_hour.values()),[])

更新

df['dh'] = df['datetime'].dt.day.map(day_hour)
df['new']=df.groupby(df['datetime'].dt.date).cumcount()

df['dh']=df.apply(lambda x : x['dh'][x['new']],axis=1)

Answer 2

我正在用你的问题打高尔夫球

sdk

熊猫：通过将二维字典映射到另一个系列来创建系列

2 个答案: