Bisiness Focused Evaluation Methods: A Case Study

Classification accuracy or other similar metrics have long been the measures used by researchers in machine learning and data mining research to compare methods and show the usefulness of such methods. Although these metrics are essential to show the predictability of the methods, they are not sufficient. In a business setting other business processes must be taken into consideration. This paper describes additional evaluations we provided potential users of our churn prediction prototype, CHAMP, to better define the characteristics of its predictions.