A Neural Network Based Customer Churn Prediction Algorithm for Telecom Sector

Abstract

For telecommunication service providers, a key method for decreasing costs and making revenue is to focus on retaining existing subscribers rather than obtaining new customers. To support this strategy, it is significant to understand customer concerns as early as possible to avoid churn. When customers switch to another competitive service provider, it results in the instant loss of business. This work focuses on building a classification model for predicting customer churn. Four different deep learning models are designed by applying different activation functions on different layers for classifying the customers into two different categories. A comparison of the performance of the different models is done by using various performance measures such as accuracy, precision, recall, and area under the curve (AUC) to determine the best activation function for the model among tanh, ReLU, ELU, and SELU.