如何从 cvs 连续数据集中计算平均值/最小值？

Question

我是 python 新手，这是我的第一个问题，如有错误请见谅。

我有一个连续测量的大 csv 文件（大约每秒测量一次，但间隔不固定）。我需要得到每分钟的平均值。我发现 groupby 可能会帮助我这样做，但我坚持将 DATE_TIME 列指定为索引和 dtype'datetime'。 csv 文件如下所示：

,DATE_TIME,N2O_dry
0,2016-03-01 02:32:02.651,0.70714453962
1,2016-03-01 02:32:03.762,0.7071444254000001
2,2016-03-01 02:32:05.257,0.70373171894
3,2016-03-01 02:32:05.953,0.70083729096
4,2016-03-01 02:32:07.049,0.69760065648
5,2016-03-01 02:32:07.928,0.6954438788699999
6,2016-03-01 02:32:08.726,0.6874527606899999
7,2016-03-01 02:32:10.005,0.6724201105500001
8,2016-03-01 02:32:10.851,0.6607286568199999
.
.
.
104503,2016-03-02 08:21:18.421,0.26879397415
104504,2016-03-02 08:21:19.532,0.26884030311
104505,2016-03-02 08:21:20.359,0.26887979686

到目前为止，我只成功地读取了数据框中的文件并将 DATE_TIME 列指定为索引，并将 DATE_TIME 列设为 dtype='datetime64[ns]' 对象：

import pandas

df=pandas.read_csv(file,usecols=[1,'N2O_dry'])
df=df.set_index('DATE_TIME')
df=pandas.to_datetime(df.index)

但是，现在我似乎只剩下 DATE_TIME 列了。有人可以帮帮我吗？

`

Answer 1

我认为您可以将参数parse_dates 和index_col 添加到read_csv，然后将resample 与mean 一起使用（这适用于pandas 0.18.0）：

import pandas as pd
import io

temp=u""",DATE_TIME,N2O_dry
0,2016-03-01 02:32:02.651,0.70714453962
1,2016-03-01 02:32:03.762,0.7071444254000001
2,2016-03-01 02:32:05.257,0.70373171894
3,2016-03-01 02:32:05.953,0.70083729096
4,2016-03-01 02:32:07.049,0.69760065648
5,2016-03-01 02:32:07.928,0.6954438788699999
6,2016-03-01 02:32:08.726,0.6874527606899999
7,2016-03-01 02:32:10.005,0.6724201105500001
8,2016-03-01 02:32:10.851,0.6607286568199999"""
#after testing replace io.StringIO(temp) to filename
df = pd.read_csv(io.StringIO(temp),
                 usecols=[1,'N2O_dry'], 
                 parse_dates=['DATE_TIME'], 
                 index_col=['DATE_TIME'])
print df
                          N2O_dry
DATE_TIME                        
2016-03-01 02:32:02.651  0.707145
2016-03-01 02:32:03.762  0.707144
2016-03-01 02:32:05.257  0.703732
2016-03-01 02:32:05.953  0.700837
2016-03-01 02:32:07.049  0.697601
2016-03-01 02:32:07.928  0.695444
2016-03-01 02:32:08.726  0.687453
2016-03-01 02:32:10.005  0.672420
2016-03-01 02:32:10.851  0.660729

print df.resample('1Min').mean()
                     N2O_dry
DATE_TIME                   
2016-03-01 02:32:00   0.6925

Answer 2

如果我理解正确，请使用

df.index = pd.to_datetime(df.index)

而不是

df = pd.to_datetime(df.index)

这应该解决问题，只剩下DATE_TIME 列。 然后你得到（在 iPython 中）：

In [27]:df.index
Out[27]: 
DatetimeIndex(['2016-03-01 02:32:02.651000', '2016-03-01 02:32:03.762000',
               '2016-03-01 02:32:05.257000', '2016-03-01 02:32:05.953000',
               '2016-03-01 02:32:07.049000', '2016-03-01 02:32:07.928000',
               '2016-03-01 02:32:08.726000', '2016-03-01 02:32:10.005000',
               '2016-03-01 02:32:10.851000'],
              dtype='datetime64[ns]', name=u'DATE_TIME', freq=None)

但还是：

In [26]: df
Out[26]: 
                          N2O_dry
DATE_TIME                        
2016-03-01 02:32:02.651  0.707145
2016-03-01 02:32:03.762  0.707144
2016-03-01 02:32:05.257  0.703732
2016-03-01 02:32:05.953  0.700837
2016-03-01 02:32:07.049  0.697601
2016-03-01 02:32:07.928  0.695444
2016-03-01 02:32:08.726  0.687453
2016-03-01 02:32:10.005  0.672420
2016-03-01 02:32:10.851  0.660729