Skip to content

evanrushton/secunda_churn

Repository files navigation

Churn Analysis for Edtech Company

Subscriber data for a small startup was analyzed to look at the retention rate, predict who will churn, and highlight features that might indicate churn or help to retain users.

A random forest model was used to produve a variable importance plot.

A logistic regressoin model was used to find coefficients for the variables and lasso regularization was used to shrink some of the parameters for the logistic regression.

An NBD/Pareto model was used to get a probability that a user will churn based on their AppOpen history. The model fit wasn't great and seasonality needs to be addressed in the time series data.

About

Predicting Churn and Subscription Insights

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages