Authors
Aliaa Salim, Fred Lacy, Yasser Ismail (Southern University and A&M College)
Abstract
This work addresses three key research questions: the spatial variability of cellular signal quality, the impact of frequency bands and network technologies (LTE vs. 5G) on performance, and the prediction of weak or unstable coverage. The study uses large-scale cellular measurement data collected via the SigCap platform across AT&T (suburban campaign), T-Mobile (urban campaign), and Verizon (Southern University campus campaign), with datasets spanning both a two-week and a two-month collection period. First, a geospatial analysis is performed to identify persistent zones of degraded coverage using signal metrics such as RSRP, RSRQ, and SINR. Next, statistical analysis is conducted to evaluate how spectrum bands and network technologies influence signal strength and quality across different environments. Finally, AI-driven predictive models, including Random Forest and gradient-based methods, are developed to identify degraded coverage using geographic and network-context features. Results show that coverage degradation is strongly location-dependent and that machine learning models can accurately predict weak or unstable regions in unseen areas. This work demonstrates the effectiveness of data-driven and AI-enabled approaches for scalable spectrum awareness, proactive coverage assessment, and next-generation network optimization.