CellDA: Data-Driven Autonomous Hotspot Cell Detection For Operational Mobile Networks

Mobile traffic demand exhibits pronounced temporal volatility, precipitating cell-level hotspots that degrade the quality of service/quality of experience (QoS/QoE) and accelerate subscriber churn. The prevailing operational practice identifies such hotspots chiefly through annual planning reviews or ad-hoc analyses triggered by customer complaints, thus providing limited temporal granularity. We reconceptualize hotspot identification as a continuous performance-management problem and propose CellDA, a scalable, data-driven framework that leverages deep learning to ingest streaming cell-level KPIs and autonomously detect incipient hotspots. Empirical evaluation on commercial LTE traces demonstrates that CellDA predicts forthcoming hotspots with high fidelity at the next time step (and substantially earlier) enabling pre-emptive mitigation. Owing to its KPI-centric design, CellDA remains access-technology-agnostic, permitting seamless deployment across 5G and prospective 6G networks via straightforward KPI mapping.