Customer churn problems can be parsed in to contractual or non-contractual. When business is contractual, you have a supervised binary classification problem. In the non-contractual setting, you have an unobserved probabilistic problem.
In this project, I modeled non-contractual customer churn in Python, using the Lifetimes implementation of BG/NBD.
Click here to view the notebook through nbviewer.
To follow along, download the notebook in this repo. Then download the data from here, and move into the same folder as the notebook. Everything should run smoothly from there.
Feel free to let me know if you have any issues, and thanks for checking out this repo!