Selective Ensemble Model for Telecom Churn Prediction

Customer based companies are concerned about costumers who decide to stop using their services (churn) because the cost of acquiring new customer is much higher than satisfying an existing customer. With the rapid development of the telecom industry, churn prediction emerges as one of the fundamental tasks for gaining the competitive advantage in the market. This paper introduces Selective Ensemble Model (SEM) as a powerful technique for churn prediction. Among a set of machine learning models, SEM dynamically selects a combination of models to participate in forming the final outcome. Experimental results show that SEM outperforms its constituent models and the averaging ensemble model.