Pandas group by timestamp and id and count

Question

I have a dataframe in the following format:

import pandas as pd
d1 = {'ID': ['A','A','A','B','B','B','B','B','C'], 
'Time': 
['1/18/2016','2/17/2016','2/16/2016','1/15/2016','2/14/2016','2/13/2016',
'1/12/2016','2/9/2016','1/11/2016'],
'Product_ID': ['2','1','1','1','1','2','1','2','2'], 
'Var_1': [0.11,0.22,0.09,0.07,0.4,0.51,0.36,0.54,0.19],
'Var_2': [1,0,1,0,1,0,1,0,1],
'Var_3': ['1','1','1','1','0','1','1','0','0']}
df1 = pd.DataFrame(d1)

Where df1 is of the form:

ID  Time        Product_ID  Var_1   Var_2   Var_3
A   1/18/2016   2           0.11    1       1
A   2/17/2016   1           0.22    0       1
A   2/16/2016   1           0.09    1       1
B   1/15/2016   1           0.07    0       1
B   2/14/2016   1           0.4     1       0
B   2/13/2016   2           0.51    0       1
B   1/12/2016   1           0.36    1       1
B   2/9/2016    2           0.54    0       0
C   1/11/2016   2           0.19    1       0

where time is in 'MM/DD/YY' format.

This is what I have to do:

1) I would like to do is to group ID's and Product ID's by Time (Specifically by each Month).

2) I want to then carry out the following column operations.
a) First, I would like to find the sum of the columns of Var_2 and Var_3 and
b) find the mean of the column Var_1.

3) Then, I would like to create a column of count of each ID and Product_ID for each month.

4) And finally, I would also like to input items ID and Product ID for which there is no entries.

For example, for ID = A and Product ID = 1 in Time = 2016-1 (January 2016), there are no observations and thus all variables take the value of 0.

Again, For ID = A and Product ID = 1 in Time = 2016-2 (January 2016),
Var_1 = (.22+.09)/2 = 0.155
Var_2 = 1,
Var_3 = 1+1=2
and finally Count = 2.

This is the output that I would like.

ID  Product_ID  Time    Var_1   Var_2   Var_3   Count
A   1           2016-1  0       0       0       0
A   1           2016-2  0.155   1       2       2
B   1           2016-1  0.215   1       1       2
B   1           2016-2  1       0.4     0       1
C   1           2016-1  0       0       0       0
C   1           2016-2  0       0       0       0
A   2           2016-1  0.11    1       1       1
A   2           2016-2  0       0       0       0
B   2           2016-1  0       0       0       0
B   2           2016-2  0.455   1       2       2
C   2           2016-1  0.19    1       0       1
C   2           2016-2  0       0       0       0

This is a little more than my programming capabilities (I know the groupby function exits but I could not figure out how to incorporate the rest of the changes). Please let me know if you have questions.

Any help will be appreciated. Thanks.

Answer 1

I break down the steps.

df1.Time=pd.to_datetime(df1.Time)
df1.Time=df1.Time.dt.month+df1.Time.dt.year*100
df1['Var_3']=df1['Var_3'].astype(int)

output=df1.groupby(['ID','Product_ID','Time']).agg({'Var_1':'mean','Var_2':'sum','Var_3':'sum'})
output=output.unstack(2).stack(dropna=False).fillna(0)# missing one .


output['Count']=output.max(1)
output.reset_index().sort_values(['Product_ID','ID'])


Out[1032]: 
  ID Product_ID    Time  Var_3  Var_2  Var_1  Count
0  A          1  201601    0.0    0.0  0.000    0.0
1  A          1  201602    2.0    1.0  0.155    2.0
4  B          1  201601    2.0    1.0  0.215    2.0
5  B          1  201602    0.0    1.0  0.400    1.0
2  A          2  201601    1.0    1.0  0.110    1.0
3  A          2  201602    0.0    0.0  0.000    0.0
6  B          2  201601    0.0    0.0  0.000    0.0
7  B          2  201602    1.0    0.0  0.525    1.0
8  C          2  201601    0.0    1.0  0.190    1.0
9  C          2  201602    0.0    0.0  0.000    0.0