Customer Churn Analysis

For the last few years Betfair is losing too many customers and is finding a solution to retain its customers. Aim of this project is to build a customer churn model to predict the customers who are about to get churned so that Betfair can implement different business strategies to retain those customers before they actually leave. The tools I am using for this analysis are R-studio and Tableau. Package mlr was chosen as the modeling package. The data for the purpose of prediction was provided by Betfair. After proper data exploration and visualization, important features for the customer churn prediction model was identified. The 8 different classification models were applied on the data in separate steps of configuring the learner task, making the learner, training the learner, prediction and performance evaluation. Out of the 8 different models, Random Forest was chosen as the best model. Cross-validation was done using random forest was done and obtained a mean miss classification error rate of 0.1278126. Hyper-parametric tuning of the random forest model was performed using package mlrHyperopt. There was only 0.05% improvement in the model accuracy after Hyper-parametric tuning. The model obtained is good enough to predict the customers who are about to fall in the churned customer category. Applying this model on the real-time data in Betfair can save huge money in revenue.