CellSense: Human Mobility Recovery via Cellular Network Data Enhancement
summary
CellSense recovers missing observations from sparse cellular billing records by combining large-scale collective patterns (mobility graph, route pruning, average travel time from a major operator’s CBR and OpenStreetMap) with context-aware individual modeling; it uses cellular signaling data as ground truth, models travel time with a Gaussian distribution and EM, and predicts locations with a bi-directional GRU, achieving a 35.3% improvement over prior methods on geographic-distance error while producing dense continuous mobility traces.
AI-written summary · about
related talks
06mExtending Coverage of Stationary Sensing Systems with Mobile Sensing Systems for Human Mobility ...
28mDiscussion: Humans As Sensors
02mData science for the most vulnerable at UNICEF
22mUsing big data from RF monitoring...
06mReal-time Travel Time Estimation with Sparse Reliable Surveillance Information
25mInsights As a Service with APIs
42mUsing Data to Understand a New Transportation Reality
32mEnriching Locations with Real-World Consumer Behavior Data