Churnsim: A Customer Churn Behavioral Simulation System for Machine Learning Education and Experimentation

Abstract

This report presents ChurnSim, a configurable simulation system that generates customer churn data for use in Machine Learning experiments and education. Customer churn behavior is of interest in a wide variety of business contexts and churn prediction with Machine Learning is now commonplace. However, raw data sources for studying customer churn are proprietary and not widely available. ChurnSim allows the generation of raw data that realistically matches a wide variety of customer behaviors related to the churn and renewal of subscription products. The data generated can be used for the study of Machine Learning churn prediction methods and for training students in applied Machine Learning methods and feature engineering.

Keywords: Simulation, Machine Learning, Churn, Customer, Subscription