04-28-2017 06:43 AM
Could you please advise which technique I could use and examples (if you have) to predict in which day/week after the last purchase, the customer is more likely to come back. I would like to have a final output like this below.
I have transactional data.., not sure what's going to be the target here.
Your help would be much appreciated .
|10003||Next 2 weeks|
|10004||Next 3 Weeks|
05-03-2017 01:03 AM
You can think of implementing two stage model here. In the first stage, you can predict the probability of a customer who is likely to return (binary classification problem that can be solved with logistic regression). If that probability is high, then in second stage, you can actually calculate the "time to return" (a continuous variable...so essentially you need to build a regression model where target variable is "time to return".
I do not have any readymade example to share. Also, i am not sure what all data is available with you. You can think of following set of features that you can include in your both models:
1. User demographics
2. User characteristics
3. Few geography specific things (e.g. store location, users' residence, how far / near the store is....assuming you have information about the registered users)
Hope this helps,