Time to event (Churn prediction)

[Niklas11]

Hello guys,

I am struggeling to built a workflow that predicts the time when someone will churn. I have a data base that consits of different Inputs Variables such as Age, time as a customer, amout of purchase and churn (yes or no) etc. Is it possible to predict the time of churn based on those characteristics if I have historical data for this (like a decision tree just with the inforomation when someone will churn not just if ? I tried to use the Kaplan Estimator for this but I thinkt this will not take my other input variables into consideration…

Thanks in advance guys

[Daniel_Weikert]

If it is not a classification problem (customer will churn / will not churn) then you can treat it as a regression problem.
So with your input features and target (“churn time” for churned customers) you should be able to use any regression algorithm you want e.g. Random Forest Regressor
br

[sanderlenselink]

Hi,

If you are familiar with Python, look at ecommerce tool:

ecommercetools

Or this one, even better if you want to forecast churn or next action:
https://lifetimes.readthedocs.io/en/latest/Quickstart.html

I don’t if it works witlin the python node of knime. Try it