0
$\begingroup$

I have a dataframe df which looks like this :

user Date TP_A TP_C TP_D TP_E TP_B TP_F Order
1 11-07-2014 0 0 1 0 0 0 0
1 11-07-2014 0 0 0 1 0 0 0
1 15-07-2014 0 0 1 0 0 0 0
1 17-07-2014 0 0 1 0 0 0 0
1 18-07-2014 0 0 1 0 0 0 0
1 25-07-2014 0 0 1 0 0 0 0
1 26-07-2014 0 0 1 0 0 0 0
1 01-08-2014 0 0 1 0 0 0 1
1 05-08-2014 0 0 1 0 0 0 0
1 12-08-2014 0 0 1 0 0 0 1
2 04-07-2014 0 0 1 0 0 0 0
2 05-07-2014 0 0 1 0 0 0 0
2 01-11-2014 0 0 1 0 0 0 1
2 09-11-2014 0 0 1 0 0 0 0
2 20-11-2014 0 0 1 0 0 0 0

I need to predict if a user's Order(target) will get executed or not and top N users whose orders are likely to get executed, based on the above training set (sample) which contains columns user_id,Date and Touch Points . Should I consider the Date to build the classification, if so how can I process the dates. The predict set has same columns but for a diferent future timeframe .

$\endgroup$
8
  • $\begingroup$ What kind of order is this ? Is there limit on number of orders in a day ? What are touch points? $\endgroup$
    – amol goel
    Aug 12, 2022 at 14:36
  • $\begingroup$ yes only one order execution per Date per user.An user can however have multiple order executions across various timeframes.You can think of touch points as actions by user, like submit form, click on link,view image.order execution(target) as purchase $\endgroup$
    – Scope
    Aug 12, 2022 at 15:50
  • $\begingroup$ One thing more . Why some orders get executed and some not ? Which industry data it is ? $\endgroup$
    – amol goel
    Aug 12, 2022 at 15:58
  • $\begingroup$ you can imagine a case of purchasing on a web app. Order execution can be treated as a buy $\endgroup$
    – Scope
    Aug 12, 2022 at 16:02
  • $\begingroup$ I can give some tips. Feature engineering on date is required like number of active days in past 10 days , whether day is weekend. Another variable you can consider is percentage of successful order in past $\endgroup$
    – amol goel
    Aug 12, 2022 at 16:11

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.