Coverage Prediction in Mobile Communication Networks: A Deep Learning Approach With a Tabular Foundation Model

Abstract

Accurate coverage prediction in mobile communication networks is crucial for optimizing performance and ensuring reliable service. However, traditional methods often struggle with the complexity and dynamic nature of wireless environments. This study introduces a novel approach leveraging a deep learning model with a tabular foundation model, TabPFN, which utilizes in-context learning and a transformer-based architecture to surpass existing techniques. Experimental validation on a real-world dataset demonstrates the model's superior prediction accuracy and adaptability, outperforming gradient boosting decision trees and supervised deep learning models in terms of root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²).