The objective was to classify churned customers for a healthcare provider who wanted to identify those customers who were not re-ordering. They also needed to create a metric to help them act accordingly which was also developed. We defined churn and quantified this as the variable to predict, ran a series of predictive models, and evaluated the models to select the best one for further interpretation and implementation.
The final result showed that Conditional Inference Tree was the most suitable model to help the client identify customers who are most likely to not re-order.
The 4th Malaysia Statistics Conference is a platform for statisticians, experts in statistics and users to discuss, exchange ideas and highlight issues pertaining to statistics for policy making and analysis.