如何根据总数的百分比拆分字段值

Question

I have sum of transactions grouped by date_month , device and channel like so

date_month   device            channel  transactions
2017-01-01  desktop         AFFILIATES           413
2017-01-01   mobile         AFFILIATES           501
2017-01-01    other         AFFILIATES            22
2017-01-01   tablet         AFFILIATES           250
2017-01-01  desktop             DIRECT         13979
etc...       etc...             etc...        etc...

date_month range is from 2017-01-01 to current date

What I'm trying to do is split the device 's other field into either mobile , desktop or tablet

Example process:

• Pivot device 'other' with its value transactions as an extra column ( other_transactions )
• Take a total sum of transactions partitioned/grouped by date_month and channel ( total_transactions )
• Then divide transactions by total_transactions to get percent total ( percent_total )
• Multiply other_transactions and percent_total to get other_split
• Add other_split to transactions to get an updated transactions field

Getting the totals and applying simple math operations shouldn't be a problem. I would do something along the lines of df['total_transactions']=df.groupby(['date_month', 'channel'])['transactions'].transform('sum') to get total_transactions but the issue I'm having is getting the other transactions into a separate column like so

date_month   device            channel  transactions  other_trans
2017-01-01  desktop         AFFILIATES           413           22
2017-01-01   mobile         AFFILIATES           501           22
2017-01-01   tablet         AFFILIATES           250           22
2017-01-01  desktop             DIRECT         13979          etc
etc...       etc...             etc...        etc...

In the end, I would like to have a data frame that removes other devices from the device column and uses its transactions to increase the remaining device transactions based on their share of transactions for that date_month and channel

Answer 1

IIUC, you can first create another dataframe using groupby , drop the rows with others , and then perform a merge :

import pandas as pd

df = pd.DataFrame({'date_month': {0: '2017-01-01', 1: '2017-01-01', 2: '2017-01-01', 3: '2017-01-01', 4: '2017-01-01', 5:"2017-01-01"},
                   'device': {0: 'desktop', 1: 'mobile', 2: 'other', 3: 'tablet', 4: 'desktop', 5:"other"},
                   'channel': {0: 'AFFILIATES', 1: 'AFFILIATES', 2: 'AFFILIATES', 3: 'AFFILIATES', 4: 'DIRECT', 5: 'DIRECT'},
                   'transactions': {0: 413, 1: 501, 2: 22, 3: 250, 4: 13979, 5: 234}})

other = df.groupby("device").get_group("other")[["date_month","channel","transactions"]]

df = df.drop(df[df["device"].str.contains("other")].index)

df = df.merge(other, on=["date_month","channel"], how="left", suffixes=("","_other"))

print (df)

Result:

   date_month   device     channel  transactions  transactions_other
0  2017-01-01  desktop  AFFILIATES           413                  22
1  2017-01-01   mobile  AFFILIATES           501                  22
2  2017-01-01   tablet  AFFILIATES           250                  22
3  2017-01-01  desktop      DIRECT         13979                 234