MIMU: Mobile WiFi Usage Inference by Mining Diverse User Behaviors
summary
MIMU predicts per-device content demand on bus WiFi to cut carrier data costs by combining a deep learning user dimension model and device dimension model that use temporal attention for regular rider schedules and spatial attention for location/line-specific patterns. Trained and evaluated on nationwide bus-WiFi logs (6,643 devices, ~150k daily active users over two months), the system drives an iterative content-caching policy, outperforms neural CF and deep ST baselines on accuracy, and yields measurable traffic and cost savings in the deployed evaluation.
AI-written summary · about
related talks
06mCellPred: A Behavior-aware Scheme for Cellular Data Usage Prediction
31mWi‑Fi/GSM Geolocation in Large Cities Using ML for 30 Million Users
07mDecentralized Attention-based Personalized Human Mobility Prediction
54mSmarter Everything with ML & the IoT
06mDeepMV: Multi-View Deep Learning for Device-Free Human Activity Recognition
30mMLDS 2019: Anubhav Shrivastava of Voot at Machine Learning Developers Summit 2019
06mPMF: A Privacy-preserving Human Mobility Prediction Framework via Federated Learning
27mPower of big data and data science in service of excellent customer experience