Predicting Customer Behaviour in the Web Hosting Industry - A Study in Mathematical Modelling

Abstract: Companies want to keep their customers. Especially, when they offer subscription based services instead of one time purchases. In the former case, if customers want to leave the company, they need to cancel their subscription. This is called customer churn. On the example of One.com, a company that offers subscription based web hosting, a mathematical model is developed to predict customer churn, so that churn preventive measures can be taken. In particular, tree based statistical learning methods such as Decision Trees and Random Forests are applied to the customer dataset of One.com and it is observed that churn predictions are made with sufficient accuracy, given that the available dataset contains information that is explanatory of churn. Then both models, Decision Tree and Random Forest, successfully deliver results that can be used for churn preventive measures on the customer base of One.com.