I'd like to build a predictive model for predicting churn for a website.

Here is the information I have for each customer :

What they did :

  • visit the website
  • Buy something
  • Do not read thenewsletter
  • Read thenewsletter

and when :

  • the date of the action

For instance for a customer X :

  • 90 days ago : visit the website
  • 50 days ago : do not read the newsetter
  • 20 days ago : do not read the newsletter

The goal is to predict if a customer is gonna be inactive based on this rule : If the customer did not visit, buy or read our email on the last 12 months, he's lost.

For now, I'm gonna use this data structure (example with the customer X) :

User_id | Last_action_1 | Last_action_2 |Last_action_3 |
  X       Do not read       Do not read       Visit

The problem is that I lost the time information, I just retain the order.

What is it recommanded to keep it ?

I think about these ideas, but I think there is better practices :

User_id | Last_action_1 |Last_action_1_date |  Last_action_2 |Last_action_2_date| Last_action_3 |Last_action_3date   
  X       Do not read      20 days ago           Do not read       50 days ago      Visit         90 days ago     
  • $\begingroup$ Last_action_1 seems to be the same as Last_action_2. This should be buy something I think. $\endgroup$ – PEV Jul 26 '13 at 13:36
  • $\begingroup$ No this user didn't be anything, take a look on his historical. $\endgroup$ – Ricol Jul 26 '13 at 14:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.