Count and cumsum per month in pandas dataframe-CodePudding

I have a pandas DataFrame with two columns: id and processing_date.

The latter being the date an item (id) was processed.

import pandas as pd
# df
  id  processed_date
 324      2016-07-08
A550      2016-07-09
  79      2016-08-10
C295      2016-08-10
 413      2016-08-11
...
 111      2021-11-08 
 709      2021-11-08

I want to plot both a graph that shows the amount of items that were processed each months and a cumulative graph "over the months". As I have 5 years and 4 months of data, I must have 64 entries and 64 data points to plot as a bar or line plot.

This is what I get from here but it's not working as expected:

df['date'] = pd.to_datetime(df['processed_date']) # needed by the nature of the data
df.set_index('date')

df = df.groupby('date')['id'].count() # <- this will stack items per day
df = df.groupby(df.index.month)['id'].count() # <- this will stack items per 12 months, but I have 5 years and 4 months of data, hence 64 different months, not 12.

How could I achieve that?

Ideal output:

# df
  nb_items_processed  cum_sum year_month
                   2        2    2016-07
                   3        5    2016-08
...
                   2      xxx    2021-11

CodePudding user response：

Start out with groupby().size() for the monthly count, then cumsum without groupby:

out = df.groupby(pd.Grouper(key='processed_date', freq='M')).size().reset_index(name='nb_items_processed')

out['cum_sum'] = out['nb_items_processed'].cumsum()