An application of the CORER classifier on customer churn prediction

Acquiring new customers in any business is much more costly than trying to retain the existing ones. So, many prediction methods have been suggested to detect churning customers. In this paper, the CORER (Colonial cOmpetitive Rule-based classifiER) classification algorithm is brought to the attention of marketing researchers to enhance the prediction accuracy of existing churn management systems. CORER is new rule-based classifier which works based on Imperialist Competitive Algorithm (ICA), a recently-proposed evolutionary optimization algorithm. Applied to the database of a telecommunication company, this classifier is found to remarkably improve accuracy in predicting churn in comparison with the most well-known techniques in the literature of the churn management, namely LOLIMOT, C5.0, neural networks and boosting classification trees. Our findings lead us to believe that the CORER classifier could cause to increase profit for the companies.