我尝试创建一个新变量，其中datetime64[ns]个对象被分配为5分钟的间隔。新的区间变量应跨越从00:00到23:55的每5分钟一段时间。赋值的标准是datetime64[ns]对象的时间是否落在相应的5分钟间隔内。我的实际数据在DateTime变量中有很多日期，但不应考虑这些不同的日期 - 只有时间元素对分配很重要。
我在下面模拟了这个。这个例子集中在大约23:30到23:45的时间段，但它应该例证我在00:00到23:55之间所有间隔所要达到的目标。我已经包含了两个随机日期来说明日期不应该有任何影响。
DateTime
2009-02-18 23:32:29 - would map to interval 23:30
2009-02-18 23:34:41 - would map to interval 23:30
2009-02-18 23:35:40 - would map to interval 23.35
2009-02-18 23:39:29 - would map to interval 23:35
2009-02-18 23:39:37 - would map to interval 23:35
2009-02-18 23:40:14 - would map to interval 23:40
2009-02-18 23:43:23 - would map to interval 23:40
2009-02-18 23:44:17 - would map to interval 23:40
...
2010-03-18 23:31:19 - also maps to interval 23:30 regardless of date
2010-03-18 23:33:31 - also maps to interval 23:30 regardless of date
2010-03-18 23:36:30 - also maps to interval 23.35 regardless of date
2010-03-18 23:38:21 - also maps to interval 23:35 regardless of date
2010-03-18 23:39:07 - also maps to interval 23:35 regardless of date
2010-03-18 23:41:44 - also maps to interval 23:40 regardless of date
2010-03-18 23:42:13 - also maps to interval 23:40 regardless of date
2010-03-18 23:43:37 - also maps to interval 23:40 regardless of date
为了清楚起见，我瞄准了这个结果：
DateTime Interval
2009-02-18 23:32:29 23:30
2009-02-18 23:34:41 23:30
2009-02-18 23:35:40 23.35
2009-02-18 23:39:29 23:35
2009-02-18 23:39:37 23:35
2009-02-18 23:40:14 23:40
2009-02-18 23:43:23 23:40
2009-02-18 23:44:17 23:40
...
2010-03-18 23:31:19 23:30
2010-03-18 23:33:31 23:30
2010-03-18 23:36:30 23.35
2010-03-18 23:38:21 23:35
2010-03-18 23:39:07 23:35
2010-03-18 23:41:44 23:40
2010-03-18 23:42:13 23:40
2010-03-18 23:43:37 23:40
我已经仔细阅读了Pandas文档，并且在这里提出了一些非常松散的相关问题，但我似乎无法获得任何可以获得正确结果的信息。
更新
这些是我的库和系统版本：
Pandas: 0.16.2
Numpy: 1.9.2
System version: '3.4.3 |Anaconda 2.3.0 (x86_64)| (default, Mar 6 2015, 12:07:41) \n[GCC 4.2.1 (Apple Inc. build 5577)]
这是完整的错误。在这里，您可以看到，根据我的实际数据，我正在使用名为datetime64[ns]的{{1}}系列。
question_time问题似乎与TypeError Traceback (most recent call last)
<ipython-input-416-d5c3256e6b40> in <module>()
----> 1 df_unique['Interval'] = ((df_unique['question_time'] - pd.TimedeltaIndex(df_unique['question_time'].dt.minute % 5, 'm')) - pd.TimedeltaIndex(df_unique['question_time'].dt.second , 's')).dt.time
//anaconda/lib/python3.4/site-packages/pandas/core/frame.py in __setitem__(self, key, value)
2125 else:
2126 # set column
-> 2127 self._set_item(key, value)
2128
2129 def _setitem_slice(self, key, value):
//anaconda/lib/python3.4/site-packages/pandas/core/frame.py in _set_item(self, key, value)
2209 # value exeption to occur first
2210 if len(self):
-> 2211 self._check_setitem_copy()
2212
2213 def insert(self, loc, column, value, allow_duplicates=False):
//anaconda/lib/python3.4/site-packages/pandas/core/generic.py in _check_setitem_copy(self, stacklevel, t, force)
1302 raise SettingWithCopyError(t)
1303 elif value == 'warn':
-> 1304 warnings.warn(t, SettingWithCopyWarning, stacklevel=stacklevel)
1305
1306 def __delitem__(self, key):
TypeError: issubclass() arg 2 must be a class or tuple of classes
有关。我尝试重置所有变量，现在我也得到了另一个操作的同样警告。
答案 0 :(得分：1)
不确定一个更好的方法，但你可以构造2个TimeDeltaIndices并从你的值中减去它，我使用模数op %来计算要减去的分钟数：
In [129]:
df['Interval'] = ((df['DateTime'] - pd.TimedeltaIndex(df['DateTime'].dt.minute % 5, 'm')) - pd.TimedeltaIndex(df['DateTime'].dt.second , 's')).dt.time
df
Out[129]:
DateTime Interval
0 2009-02-18 23:32:29 23:30:00
1 2009-02-18 23:34:41 23:30:00
2 2009-02-18 23:35:40 23:35:00
3 2009-02-18 23:39:29 23:35:00
4 2009-02-18 23:39:37 23:35:00
5 2009-02-18 23:40:14 23:40:00
6 2009-02-18 23:43:23 23:40:00
7 2009-02-18 23:44:17 23:40:00
8 2010-03-18 23:31:19 23:30:00
9 2010-03-18 23:33:31 23:30:00
10 2010-03-18 23:36:30 23:35:00
11 2010-03-18 23:38:21 23:35:00
12 2010-03-18 23:39:07 23:35:00
13 2010-03-18 23:41:44 23:40:00
14 2010-03-18 23:42:13 23:40:00
15 2010-03-18 23:43:37 23:40:00