This is a pretty classic binary classification problem framed in the context of a real-world scenario (and is actually a problem that many subscription-based businesses face).
The problem statement for this project would be something like:
Given a specific customer’s history, how might we predict if that specific customer will leave us soon?
In a real application, this process would be automated and run in the background. But since I don’t have a business to apply this to, and since just showing the code is pretty ineffective, I built a web interface that allows you to modify the various parameters and tweak the results.
I did not contribute much to the creation of this model - it’s taken mostly from chapters 3-6 of Machine Learning Book Camp by Alexey Grigorev. However, I did work to take the initial Flask-based API interface and built out a Django application and associated webview.
And a final ironic twist. When doing initial data analysis, I was (and currently still am) living on a farm and forced to use a hotspot for internet while I wait on fiber optic internet.
Why is that ironic?
One of the biggest predictive values that a customer will churn in the dataset is if the customer is a fiber optic subscriber. And I really want to churn from my hotspot to become a fiber customer