Deep Neural Network Model for Customer Attrition Forecast in a Telecommunication Company

ABSTRACT The loss of customers is becoming a significant challenge for telecom companies due to the high cost of acquiring new customers and the critical need to retain existing ones. This dissertation explores the importance of predicting customer attrition in the telecommunications sector using a deep neural network (DNN) model. The study highlights the crucial role of customer retention in a highly competitive market. The system was developed using historical data, preprocessing techniques, and a customized DNN architecture. The methodology followed a DevOps approach, encompassing the collection, integration, and preprocessing of diverse datasets, followed by the construction and optimization of the DNN model with five layers using stochastic gradient descent. The findings demonstrate the model’s impressive accuracy, achieving 98.1% after 100 epochs, along with improved precision. The results underscore the DNN model's effectiveness in predicting churn, emphasizing the value of iterative refinement through multiple training cycles. This research offers valuable insights and practical methodologies for telecom companies aiming to adopt proactive strategies to enhance customer retention and satisfaction in a dynamic and competitive environment. Keywords: Churn, DNN, Dataset, Epochs, Prediction, Optimization.