Pandas Dataframe的时频变化

Question

我有一个熊猫DataFrame如下所示。

df
                              A           B
date_time                                    
2014-07-01 06:03:59.614000  62.1250       NaN
2014-07-01 06:03:59.692000  62.2500       NaN
2014-07-01 06:13:34.524000  62.2500  241.0625
2014-07-01 06:13:34.602000  62.2500  241.5000
2014-07-01 06:15:05.399000  62.2500  241.3750
2014-07-01 06:15:05.399000  62.2500  241.2500
2014-07-01 06:15:42.004000  62.2375  241.2500
2014-07-01 06:15:42.082000  62.2375  241.3750
2014-07-01 06:15:42.082000  62.2375  240.2500

我想把这个频率改变为有规律的1 minute间隔。但是，请在下面获得错误：

new = df.asfreq('1Min')
>>error: cannot reindex from a duplicate axis

现在，我明白为什么会这样了。由于我的时间粒度很高(以毫秒为单位)，但不规则，所以我每分钟获得多个读数，甚至每秒都是如此。所以我试着把这些毫秒读数和几分钟结合起来，然后去掉下面的复制件。

# try to convert the index to minutes and drop duplicates
df['index'] = df.index
df['minute_index']= df['index'].apply( lambda x: x.strftime('%Y-%m-%d %H:%M'))
df.drop_duplicates(cols = 'minute_index', inplace = True, take_last = True)
df_by_minute = df.set_index('minute_index')
df_by_minute
                        A                B               index
minute_index                                                     
2014-07-01 06:03    62.2500        NaN 2014-07-01 06:03:59.692000
2014-07-01 06:13    62.2500     241.50 2014-07-01 06:13:34.602000
2014-07-01 06:15    62.2375     240.25 2014-07-01 06:15:42.082000

# now change the frequency to 1 minute but I just get NaNs (!)
df_by_minute.asfreq('1Min')
                            A          B   index
2014-07-01 06:03:00        NaN        NaN   NaT
2014-07-01 06:04:00        NaN        NaN   NaT
2014-07-01 06:05:00        NaN        NaN   NaT
2014-07-01 06:06:00        NaN        NaN   NaT
2014-07-01 06:07:00        NaN        NaN   NaT
2014-07-01 06:08:00        NaN        NaN   NaT
2014-07-01 06:09:00        NaN        NaN   NaT
2014-07-01 06:10:00        NaN        NaN   NaT
2014-07-01 06:11:00        NaN        NaN   NaT
2014-07-01 06:12:00        NaN        NaN   NaT
2014-07-01 06:13:00        NaN        NaN   NaT
2014-07-01 06:14:00        NaN        NaN   NaT
2014-07-01 06:15:00        NaN        NaN   NaT

如你所见，它不起作用。有人能帮忙吗？我试图实现的是获得一个返回A or B as of DateTime和DateTime的函数，该函数将以1分钟递增。

Answer 1

我认为，不是asfreq，而是resample适合您的需要：

new = df.resample('T', how='mean')

对于how选项，也可以使用“最后”或“第一”。

Answer 2

变换数据帧的Pandas Dataframe.resample()函数具有将时间频率从秒变到分钟、小时、日、年等功能，它与DatetimeIndex字段和datetime列完美地工作。

​

​

​

​

​

​