Customer switching behavior analysis in the telecommunication industry via push-pull-mooring framework: a machine learning approach

Abstract Customer retention is one of the key challenges in the telecommunication industry. Companies may find customer churn prediction to be vital to the success of their operations because a careful analysis of churning may provide a crucial means to retain customers. Among potentially a vast amount of factors that impact the churn, it is critical to identify the most influential ones towards which customer retention efforts can be directed. In this paper, we compare the performance of different churn prediction models based on the real data obtained from a partner company. The prediction models include logistic regression, support vector machines, random forest, and decision tree. Furthermore, the push-pull-mooring (PPM) framework is utilized to study the effect of features on customers churn behavior from push, pull, and mooring perspectives. A partial least squares (PLS) regression is used to perform the PPM analysis. Furthermore, the behavior or churners and non-churners are analyzed. The results show that the logistic regression has the highest prediction accuracy, and the drop-percentage push factor is found to be one of the most influential factors affecting the customer churn, as churners are more sensitive to service quality than non-churners.