Deriving Real-time City Crowd Flows by Heterogeneous Big Urban Data

Bo Tang, Chuan Yang, Long Xiang, Jian Zeng

Abstract: Real-time city crowd flows are extremely important for facility placement, transportation management, and public safety. In this paper, we show how to derive the real-time city crowd flows with heterogeneous urban data. Unlike existing prediction-based approaches, our proposal does not rely on the training data and learning models. We propose a computation framework for it by exploiting the massive heterogeneous urban data, which includes both immutable data (i.e., bus and subway stations, commercial-based regions) and mutable data (i.e., real-time taxi, bus, subway transactions and trajectories). In addition, our solution provides accurate and timely city crowd-flows.To provide a practical solution for it, we first partition the city into commercial-based regions with geographic information (e.g., road network, administrative regions). Then, we devise three major components (i.e., urban data fusion model, heterogeneous urban data integration model, and effective crowd flows computation model) in the computation framework to process massive heterogeneous urban data effectively and derive city crowd flows accurately. Finally, we conduct extensive experiments to demonstrate the effectiveness and efficiency of our proposed solution with real heterogeneous urban data in Shenzhen, China.