Rough Set Decision Rules for Usage-Based Churn Modeling in Mobile Telecommunications

The concern of customer churn significantly impacts the telecommunications industry, given the considerable costs associated with acquiring new customers compared to retaining existing ones. To effectively guide anti-churn initiatives, it becomes crucial to point profitable clients with the highest likelihood of churning. However, the data utilized for identifying potential churners often carries inherent imprecision. In this study, we use a rough sets modeling approach for discerning churn intent based on usage data within the mobile telecommunications domain. This approach takes into account the uncertainty in the data and provides concise, easy-to-interpret rules. In the paper we use four approaches: exhaustive algorithm, genetic algorithm, covering algorithm and LEM2 algorithm and attribute discretization method. Finally, it was found that the best results are obtained using the rough set rule-based systems with the LEM2 algorithm and attribute discretization. A very high accuracy of 0.997 was achieved. The results for the analyzed case are better than those generated by a fuzzy rule-based system.