CellPred: A Behavior-aware Scheme for Cellular Data Usage Prediction
summary
Claims per-user cellular data prediction improves tower-level forecasts by adding behavioral tags from subscription records and mobility/usage signals. Uses a soft attention encoder-decoder that fuses mobility traces, per-interval usage, and explicit user tags, trained on signaling logs from 3 million phones and aggregated to 23,000 cell towers. Reports lower MAPE than statistical baselines, especially in suburban cells and on weekends, and evaluates ablations removing mobility or usage features.
AI-written summary · about
related talks
06mMIMU: Mobile WiFi Usage Inference by Mining Diverse User Behaviors
07mHERMAS: A Human Mobility Embedding Framework with Large-scale Cellular Signaling Data
44mFrom responsive to anticipatory
50mHow to build propensity model where churn is single digit
19mHow to Leverage Predictive Analytics to Enhance iOS Marketing Performance
07mMIFF: Human Mobility Extractions with Cellular Signaling Data under Spatio-temporal Uncertainty
21m“Trust Us”: Mobile Phone Use Patterns Can Predict Individual Trust Propensity
21mFrom Reactive to Proactive: How AI is Redefining Resilience in Modern Networks