Python:在熊猫中计算它们之后对日期进行排序

时间:2017-03-03 08:50:08

标签: python list pandas datetime frequency-distribution

我从字符串中提取日期和时间,然后通过wrintig将它们转换为Pandas DatFrame:

df = pd.to_datetime(news_date, format='%m/%d/%Y')

,输出如下:

['1997-10-31 18:00:00', '1997-10-31 18:00:00',
           '1997-10-31 18:00:00', '1997-10-31 18:00:00',
           '1997-10-31 18:00:00', '1997-10-31 18:00:00',
           '1997-10-31 18:00:00', '1997-10-31 18:00:00',
           '1997-10-31 18:00:00', '1997-10-31 18:00:00',
           ...
           '2016-12-07 03:14:00', '2016-12-09 16:31:00',
           '2016-12-10 19:02:00', '2016-12-11 09:41:00',
           '2016-12-12 05:01:00', '2016-12-12 05:39:00',
           '2016-12-12 06:44:00', '2016-12-12 08:11:00',
           '2016-12-12 09:36:00', '2016-12-12 10:19:00']

然后我想只保留月份和年份并对日期进行排序,我写道:

month_year = df.to_series().apply(lambda x: dt.datetime.strftime(x, '%m-%Y')).tolist() # remove time and day
new = sorted(month_year, key=lambda x: datetime.datetime.strptime(x, '%m-%Y'))  # sort date

到目前为止,我有一份日期清单。当我尝试计算它们的频率时会出现问题(我必须稍后绘制时间分布)。 我的代码是:

print(pd.DataFrame(new).groupby(month_year).count())

,输出为:

01-1998   60
01-1999   18
01-2000   49
01-2001   50
01-2002   87
01-2003  129
01-2004  125
01-2005  225
01-2006  154
01-2007  302
01-2008  161
01-2009  161
01-2010  167
01-2011  181
01-2012  227
...      ...
12-2014   82
12-2015   89
12-2016   13

尽管如此,我希望在一列中有一个排序日期,并且它的频率在另一列(例如,Pandas DataFrame)中可以轻松绘制,例如:

01-1998    60
02-1998    32
03-1998    22
...       ...
11-2016    20
12-2016    13

1 个答案:

答案 0 :(得分:2)

我认为您需要转换to_period然后转换value_counts month period才能进行排序sort_index

news_date = ['1997-10-31 18:00:00', '1997-10-31 18:00:00',
           '1997-10-30 18:00:00', '1997-10-30 18:00:00',
           '1997-10-30 18:00:00', '1997-10-30 18:00:00',
           '1997-11-30 18:00:00', '1997-11-30 18:00:00',
           '1997-12-30 18:00:00', '1997-12-30 18:00:00',
           '2016-12-07 03:14:00', '2016-01-09 16:31:00',
           '2016-12-10 19:02:00', '2016-01-11 09:41:00',
           '2016-12-12 05:01:00', '2016-02-12 05:39:00',
           '2016-12-12 06:44:00', '2016-12-12 08:11:00',
           '2016-12-12 09:36:00', '2016-12-12 10:19:00']

idx = pd.to_datetime(news_date)
new = pd.Series(idx.to_period('m'))
print (new)
0    1997-10
1    1997-10
2    1997-10
3    1997-10
4    1997-10
5    1997-10
6    1997-11
7    1997-11
8    1997-12
9    1997-12
10   2016-12
11   2016-01
12   2016-12
13   2016-01
14   2016-12
15   2016-02
16   2016-12
17   2016-12
18   2016-12
19   2016-12
dtype: object
df = new.value_counts().sort_index().reset_index()
df.columns = ['Date','Count']
df.Date = df.Date.dt.strftime('%Y-%m')
print (df)
      Date  Count
0  1997-10      6
1  1997-11      2
2  1997-12      2
3  2016-01      2
4  2016-02      1
5  2016-12      7

另一种可能的解决方案是先strftime转换为strings

new = pd.Series(idx.strftime('%Y-%m'))
df = new.value_counts().sort_index().reset_index()
df.columns = ['Date','Count']
print (df)
     Date  Count
0  1997-10      6
1  1997-11      2
2  1997-12      2
3  2016-01      2
4  2016-02      1
5  2016-12      7