White Paper
NN-Based Predictive Analytics for Prepaid Mobile Services in Telco Sector
This whitepaper studies neural-network-based predictive analytics for prepaid mobile services. It focuses on predicting whether customers will top up within two, three, or four days and estimating account balance before top-up events. The goal is to help telecom operators time marketing actions more accurately and personalize campaigns. The paper compares neural networks with traditional production models such as Random Forest and Gradient Boosting Trees, using preprocessing, feature selection, hyperparameter tuning, and validation. Results show neural networks can significantly improve customer behavior prediction.
