Customer Attrition Forecast Model in Telecommunication Company using Deep Artificial Neural Network (DNN)

ABSTRACT Customer attrition presents a significant challenge for telecommunications companies due to the high cost of acquiring new customers and the critical importance of retaining existing ones. This study investigates the use of a Deep Neural Network (DNN) model to predict customer churn, emphasizing its strategic role in improving customer retention in a highly competitive market. The system was developed using a publicly available dataset from Kaggle.com, comprising 3,333 instances and 20 features, and included extensive data preprocessing and a customized DNN architecture.The methodology follows a DevOps-driven machine learning pipeline encompassing data collection, preprocessing, model development, training, and evaluation. Key reprocessing steps included feature selection, normalization, and encoding of categorical variables to prepare the data for modeling. The DNN architecture was designed to handle the high-dimensional feature space, employing multiple hidden layers with ReLU activation functions and dropout regularization to mitigate overfitting. The model was trained using stochastic gradient descent (SGD) with adaptive learning rates. Hyper-parameters such as learning rate, batch size, and number of neurons were optimized through experimentation and k-fold cross-validation.The model achieved a classification accuracy of 98.1% after 100 training epochs, along with notable improvements in precision. These results highlight the efficacy of the proposed DNN model in predicting customer churn and underscore the benefits of iterative model refinement. This research provides actionable insights for telecom providers seeking to implement predictive analytics for proactive customer retention strategies in a dynamic market landscape. Keywords: Churn, DNN, Dataset, Epochs, Prediction, Optimization.