如何根据特定列表计算频率？

Question

I have a DataFrame that looks like this.

                date name
0 2015-06-13 00:21:25    a
1 2015-06-13 01:00:25    b
2 2015-06-13 02:54:48    c
3 2015-06-15 14:38:15    a
4 2015-06-15 15:29:28    b

I want to count the occurrences of dates against a specific date range, including ones that do not appear in the column (and ignores whatever that is in the name column). For example, I might have a date range that looks like this:

periods = pd.date_range('2015-06-13', '2015-06-16', freq = 'd')

Then, I want an output that looks something like:

date       count    
2015-06-13 3
2015-06-14 0
2015-06-15 2
2015-06-16 0

I haven't been able to find any function that let me keep the 0 rows.

Answer 1

I think you can first use date from column date for value_counts and then reindex by periods with fillna by 0 . Last convert float to int by astype and reset_index :

df = df['date'].dt.date.value_counts()
print df
2015-06-13    3
2015-06-15    2
Name: date, dtype: int64

periods = pd.date_range('2015-06-13', '2015-06-16', freq = 'd')

df = df.reindex(periods).fillna(0).astype(int).reset_index()
df.columns = ['date','count']
print df
        date  count
0 2015-06-13      3
1 2015-06-14      0
2 2015-06-15      2
3 2015-06-16      0

Answer 2

This is very similar to the solution of @jezrael, but uses a groupby instead of value_counts:

>>> (pd.DataFrame(df.groupby(df.date.dt.date)['name']
                    .count()
                    .reindex(periods)
                    .fillna(0))
     .rename(columns={'name': 'count'}))
            count
2015-06-13      3
2015-06-14      0
2015-06-15      2
2015-06-16      0

Note: In Pandas 0.18.0 the reindex operation changes the type of count from ints to floats, so if you are using that version you'll need to tack on .astype(int) to the end.