Big data-based urban crowd flow prediction approaches

Given the need to decrease traffic accidents and the corresponding human and socials cost which are resulted from not only inadequate driving, but also from the inaccurate planning of the flow conditions, thus, crowd flow prediction is a fundamental urban computing problem. Several approaches have been suggested to solve this problem some of which rely on big data and deep learning. Big data and deep learning have been successfully applied in different studies. Besides, having adequate data is usually a prerequisite for such purposes, especially when big data and deep learning are adopted. Hence, present paper critically reviews and analyzes crowd flow prediction approaches. This survey initially provides fundamentals of crowd analysis including crowd video analysis and big data crowd analysis and then focuses on big data analysis. For this purpose, we highlight three main works conducted accordingly which include Spatio -Temporal Residual Networks or ST-ResNet, Deep Spatio-Temporal Transfer Learning and Citywide Crowd Flows (FCCF).

Big data-based urban crowd flow prediction approaches Keywords:

urban crowd flow prediction , video analysis , big-data , deep learning